diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -11,3909 +11,3909 @@
   "log_history": [
     {
       "epoch": 2e-05,
-      "grad_norm": 2.7753920555114746,
+      "grad_norm": 2.618687152862549,
       "learning_rate": 0.0,
-      "loss": 1.7812,
+      "loss": 1.7883,
       "step": 1
     },
     {
       "epoch": 0.002,
-      "grad_norm": 2.502140522003174,
+      "grad_norm": 2.935516357421875,
       "learning_rate": 9.9e-07,
-      "loss": 1.7681,
+      "loss": 1.7832,
       "step": 100
     },
     {
       "epoch": 0.004,
-      "grad_norm": 2.7424938678741455,
+      "grad_norm": 2.6478724479675293,
       "learning_rate": 1.9900000000000004e-06,
-      "loss": 1.7633,
+      "loss": 1.8052,
       "step": 200
     },
     {
       "epoch": 0.006,
-      "grad_norm": 2.4527595043182373,
+      "grad_norm": 2.5077390670776367,
       "learning_rate": 2.99e-06,
-      "loss": 1.7876,
+      "loss": 1.8003,
       "step": 300
     },
     {
       "epoch": 0.008,
-      "grad_norm": 2.787921667098999,
+      "grad_norm": 2.803349733352661,
       "learning_rate": 3.990000000000001e-06,
-      "loss": 1.7655,
+      "loss": 1.7773,
       "step": 400
     },
     {
       "epoch": 0.01,
-      "grad_norm": 2.5596084594726562,
+      "grad_norm": 2.7017805576324463,
       "learning_rate": 4.9900000000000005e-06,
-      "loss": 1.7712,
+      "loss": 1.7897,
       "step": 500
     },
     {
       "epoch": 0.012,
-      "grad_norm": 2.5325963497161865,
+      "grad_norm": 2.8483221530914307,
       "learning_rate": 4.999950652140343e-06,
-      "loss": 1.7384,
+      "loss": 1.7942,
       "step": 600
     },
     {
       "epoch": 0.014,
-      "grad_norm": 2.3152921199798584,
+      "grad_norm": 2.7330849170684814,
       "learning_rate": 4.999800611667942e-06,
-      "loss": 1.7508,
+      "loss": 1.789,
       "step": 700
     },
     {
       "epoch": 0.016,
-      "grad_norm": 2.4681248664855957,
+      "grad_norm": 2.806673526763916,
       "learning_rate": 4.9995498795914055e-06,
-      "loss": 1.7415,
+      "loss": 1.8014,
       "step": 800
     },
     {
       "epoch": 0.018,
-      "grad_norm": 2.457768678665161,
+      "grad_norm": 2.6345088481903076,
       "learning_rate": 4.999198466010187e-06,
-      "loss": 1.7537,
+      "loss": 1.7852,
       "step": 900
     },
     {
       "epoch": 0.02,
-      "grad_norm": 2.54608416557312,
+      "grad_norm": 2.9260714054107666,
       "learning_rate": 4.998746385079172e-06,
-      "loss": 1.7411,
+      "loss": 1.7966,
       "step": 1000
     },
     {
       "epoch": 0.02,
-      "eval_loss": 1.7543565034866333,
-      "eval_runtime": 30.099,
-      "eval_samples_per_second": 166.118,
-      "eval_steps_per_second": 2.625,
+      "eval_loss": 1.787523865699768,
+      "eval_runtime": 30.0699,
+      "eval_samples_per_second": 166.279,
+      "eval_steps_per_second": 2.627,
       "step": 1000
     },
     {
       "epoch": 0.022,
-      "grad_norm": 2.396301746368408,
+      "grad_norm": 2.851686716079712,
       "learning_rate": 4.9981936550081165e-06,
-      "loss": 1.7197,
+      "loss": 1.7904,
       "step": 1100
     },
     {
       "epoch": 0.024,
-      "grad_norm": 2.6680235862731934,
+      "grad_norm": 2.5083374977111816,
       "learning_rate": 4.997540298060907e-06,
-      "loss": 1.7378,
+      "loss": 1.7859,
       "step": 1200
     },
     {
       "epoch": 0.026,
-      "grad_norm": 2.520023822784424,
+      "grad_norm": 2.4664571285247803,
       "learning_rate": 4.9967863405546644e-06,
-      "loss": 1.7338,
+      "loss": 1.7997,
       "step": 1300
     },
     {
       "epoch": 0.028,
-      "grad_norm": 2.5510494709014893,
+      "grad_norm": 2.637437582015991,
       "learning_rate": 4.995931812858689e-06,
-      "loss": 1.7371,
+      "loss": 1.7967,
       "step": 1400
     },
     {
       "epoch": 0.03,
-      "grad_norm": 2.7541258335113525,
+      "grad_norm": 2.8107752799987793,
       "learning_rate": 4.99497674939323e-06,
-      "loss": 1.7279,
+      "loss": 1.7594,
       "step": 1500
     },
     {
       "epoch": 0.032,
-      "grad_norm": 2.751742124557495,
+      "grad_norm": 2.616604804992676,
       "learning_rate": 4.993921188628106e-06,
-      "loss": 1.714,
+      "loss": 1.7775,
       "step": 1600
     },
     {
       "epoch": 0.034,
-      "grad_norm": 2.5539989471435547,
+      "grad_norm": 2.7829654216766357,
       "learning_rate": 4.99276517308115e-06,
-      "loss": 1.7324,
+      "loss": 1.778,
       "step": 1700
     },
     {
       "epoch": 0.036,
-      "grad_norm": 2.7032289505004883,
+      "grad_norm": 2.825524091720581,
       "learning_rate": 4.9915087493165e-06,
-      "loss": 1.7199,
+      "loss": 1.774,
       "step": 1800
     },
     {
       "epoch": 0.038,
-      "grad_norm": 2.71175479888916,
+      "grad_norm": 2.745628595352173,
       "learning_rate": 4.9901519679427214e-06,
-      "loss": 1.7318,
+      "loss": 1.7847,
       "step": 1900
     },
     {
       "epoch": 0.04,
-      "grad_norm": 2.5806422233581543,
+      "grad_norm": 2.7032053470611572,
       "learning_rate": 4.98869488361077e-06,
-      "loss": 1.7194,
+      "loss": 1.7966,
       "step": 2000
     },
     {
       "epoch": 0.04,
-      "eval_loss": 1.7078638076782227,
-      "eval_runtime": 24.9549,
-      "eval_samples_per_second": 200.362,
-      "eval_steps_per_second": 3.166,
+      "eval_loss": 1.8017730712890625,
+      "eval_runtime": 24.7305,
+      "eval_samples_per_second": 202.18,
+      "eval_steps_per_second": 3.194,
       "step": 2000
     },
     {
       "epoch": 0.042,
-      "grad_norm": 2.499134063720703,
+      "grad_norm": 2.6376078128814697,
       "learning_rate": 4.987137555011793e-06,
-      "loss": 1.7176,
+      "loss": 1.7801,
       "step": 2100
     },
     {
       "epoch": 0.044,
-      "grad_norm": 2.601299285888672,
+      "grad_norm": 2.885993003845215,
       "learning_rate": 4.9854800448747555e-06,
-      "loss": 1.7326,
+      "loss": 1.7852,
       "step": 2200
     },
     {
       "epoch": 0.046,
-      "grad_norm": 2.46392560005188,
+      "grad_norm": 2.670661449432373,
       "learning_rate": 4.983722419963926e-06,
-      "loss": 1.7409,
+      "loss": 1.7811,
       "step": 2300
     },
     {
       "epoch": 0.048,
-      "grad_norm": 2.7927350997924805,
+      "grad_norm": 2.727155923843384,
       "learning_rate": 4.98186475107618e-06,
-      "loss": 1.738,
+      "loss": 1.7935,
       "step": 2400
     },
     {
       "epoch": 0.05,
-      "grad_norm": 2.562814712524414,
+      "grad_norm": 2.6564817428588867,
       "learning_rate": 4.979907113038148e-06,
-      "loss": 1.7247,
+      "loss": 1.7725,
       "step": 2500
     },
     {
       "epoch": 0.052,
-      "grad_norm": 2.4681177139282227,
+      "grad_norm": 2.7066261768341064,
       "learning_rate": 4.977849584703206e-06,
-      "loss": 1.7258,
+      "loss": 1.7946,
       "step": 2600
     },
     {
       "epoch": 0.054,
-      "grad_norm": 2.479094982147217,
+      "grad_norm": 2.6405227184295654,
       "learning_rate": 4.975692248948294e-06,
-      "loss": 1.731,
+      "loss": 1.7991,
       "step": 2700
     },
     {
       "epoch": 0.056,
-      "grad_norm": 2.499499559402466,
+      "grad_norm": 2.9687182903289795,
       "learning_rate": 4.9734351926705836e-06,
-      "loss": 1.7189,
+      "loss": 1.7876,
       "step": 2800
     },
     {
       "epoch": 0.058,
-      "grad_norm": 2.721623182296753,
+      "grad_norm": 2.5749998092651367,
       "learning_rate": 4.971078506783971e-06,
-      "loss": 1.717,
+      "loss": 1.7964,
       "step": 2900
     },
     {
       "epoch": 0.06,
-      "grad_norm": 2.681560516357422,
+      "grad_norm": 2.8419148921966553,
       "learning_rate": 4.968622286215421e-06,
-      "loss": 1.7214,
+      "loss": 1.7846,
       "step": 3000
     },
     {
       "epoch": 0.06,
-      "eval_loss": 1.685670256614685,
-      "eval_runtime": 24.1266,
-      "eval_samples_per_second": 207.24,
-      "eval_steps_per_second": 3.274,
+      "eval_loss": 1.7648264169692993,
+      "eval_runtime": 24.7981,
+      "eval_samples_per_second": 201.628,
+      "eval_steps_per_second": 3.186,
       "step": 3000
     },
     {
       "epoch": 0.062,
-      "grad_norm": 2.6745054721832275,
+      "grad_norm": 2.798104763031006,
       "learning_rate": 4.966066629901139e-06,
-      "loss": 1.7133,
+      "loss": 1.787,
       "step": 3100
     },
     {
       "epoch": 0.064,
-      "grad_norm": 2.5229477882385254,
+      "grad_norm": 2.518934965133667,
       "learning_rate": 4.963411640782589e-06,
-      "loss": 1.711,
+      "loss": 1.7843,
       "step": 3200
     },
     {
       "epoch": 0.066,
-      "grad_norm": 2.5905637741088867,
+      "grad_norm": 2.6785619258880615,
       "learning_rate": 4.960657425802345e-06,
-      "loss": 1.6994,
+      "loss": 1.7811,
       "step": 3300
     },
     {
       "epoch": 0.068,
-      "grad_norm": 2.480076313018799,
+      "grad_norm": 2.6914544105529785,
       "learning_rate": 4.9578040958997865e-06,
-      "loss": 1.7211,
+      "loss": 1.7699,
       "step": 3400
     },
     {
       "epoch": 0.07,
-      "grad_norm": 2.4334428310394287,
+      "grad_norm": 2.656628131866455,
       "learning_rate": 4.954851766006624e-06,
-      "loss": 1.6982,
+      "loss": 1.7758,
       "step": 3500
     },
     {
       "epoch": 0.072,
-      "grad_norm": 2.7318227291107178,
+      "grad_norm": 2.8454766273498535,
       "learning_rate": 4.951800555042277e-06,
-      "loss": 1.7088,
+      "loss": 1.7588,
       "step": 3600
     },
     {
       "epoch": 0.074,
-      "grad_norm": 3.27542781829834,
+      "grad_norm": 2.551102876663208,
       "learning_rate": 4.948650585909078e-06,
-      "loss": 1.7124,
+      "loss": 1.7608,
       "step": 3700
     },
     {
       "epoch": 0.076,
-      "grad_norm": 2.6027004718780518,
+      "grad_norm": 2.777878999710083,
       "learning_rate": 4.945401985487322e-06,
-      "loss": 1.7179,
+      "loss": 1.7898,
       "step": 3800
     },
     {
       "epoch": 0.078,
-      "grad_norm": 2.4319632053375244,
+      "grad_norm": 2.7975049018859863,
       "learning_rate": 4.942054884630163e-06,
-      "loss": 1.7232,
+      "loss": 1.7884,
       "step": 3900
     },
     {
       "epoch": 0.08,
-      "grad_norm": 2.408602476119995,
+      "grad_norm": 2.484600067138672,
       "learning_rate": 4.938609418158333e-06,
-      "loss": 1.7023,
+      "loss": 1.7821,
       "step": 4000
     },
     {
       "epoch": 0.08,
-      "eval_loss": 1.7144120931625366,
-      "eval_runtime": 24.0336,
-      "eval_samples_per_second": 208.042,
-      "eval_steps_per_second": 3.287,
+      "eval_loss": 1.7807953357696533,
+      "eval_runtime": 23.8739,
+      "eval_samples_per_second": 209.434,
+      "eval_steps_per_second": 3.309,
       "step": 4000
     },
     {
       "epoch": 0.082,
-      "grad_norm": 2.8398849964141846,
+      "grad_norm": 2.7985785007476807,
       "learning_rate": 4.9350657248547225e-06,
-      "loss": 1.7175,
+      "loss": 1.7798,
       "step": 4100
     },
     {
       "epoch": 0.084,
-      "grad_norm": 2.5136523246765137,
+      "grad_norm": 2.7834692001342773,
       "learning_rate": 4.931423947458778e-06,
-      "loss": 1.7092,
+      "loss": 1.7829,
       "step": 4200
     },
     {
       "epoch": 0.086,
-      "grad_norm": 2.67118501663208,
+      "grad_norm": 2.577984094619751,
       "learning_rate": 4.9276842326607634e-06,
-      "loss": 1.7009,
+      "loss": 1.7696,
       "step": 4300
     },
     {
       "epoch": 0.088,
-      "grad_norm": 2.8670244216918945,
+      "grad_norm": 2.582216262817383,
       "learning_rate": 4.923846731095845e-06,
-      "loss": 1.6965,
+      "loss": 1.7796,
       "step": 4400
     },
     {
       "epoch": 0.09,
-      "grad_norm": 2.7709720134735107,
+      "grad_norm": 2.5914933681488037,
       "learning_rate": 4.919911597338026e-06,
-      "loss": 1.72,
+      "loss": 1.7899,
       "step": 4500
     },
     {
       "epoch": 0.092,
-      "grad_norm": 2.672072649002075,
+      "grad_norm": 2.629214286804199,
       "learning_rate": 4.915878989893922e-06,
-      "loss": 1.704,
+      "loss": 1.7807,
       "step": 4600
     },
     {
       "epoch": 0.094,
-      "grad_norm": 2.786982774734497,
+      "grad_norm": 2.5505175590515137,
       "learning_rate": 4.91174907119637e-06,
-      "loss": 1.6832,
+      "loss": 1.7555,
       "step": 4700
     },
     {
       "epoch": 0.096,
-      "grad_norm": 2.453965663909912,
+      "grad_norm": 2.7634997367858887,
       "learning_rate": 4.907522007597896e-06,
-      "loss": 1.7192,
+      "loss": 1.7586,
       "step": 4800
     },
     {
       "epoch": 0.098,
-      "grad_norm": 2.6247568130493164,
+      "grad_norm": 2.897411823272705,
       "learning_rate": 4.903197969364003e-06,
-      "loss": 1.6942,
+      "loss": 1.7997,
       "step": 4900
     },
     {
       "epoch": 0.1,
-      "grad_norm": 2.6226806640625,
+      "grad_norm": 2.698044538497925,
       "learning_rate": 4.898777130666322e-06,
-      "loss": 1.6912,
+      "loss": 1.7721,
       "step": 5000
     },
     {
       "epoch": 0.1,
-      "eval_loss": 1.7000234127044678,
-      "eval_runtime": 24.0689,
-      "eval_samples_per_second": 207.737,
-      "eval_steps_per_second": 3.282,
+      "eval_loss": 1.7994335889816284,
+      "eval_runtime": 23.8937,
+      "eval_samples_per_second": 209.26,
+      "eval_steps_per_second": 3.306,
       "step": 5000
     },
     {
       "epoch": 0.102,
-      "grad_norm": 2.4851531982421875,
+      "grad_norm": 2.906372308731079,
       "learning_rate": 4.89425966957559e-06,
-      "loss": 1.7016,
+      "loss": 1.7841,
       "step": 5100
     },
     {
       "epoch": 0.104,
-      "grad_norm": 2.585747003555298,
+      "grad_norm": 2.4340624809265137,
       "learning_rate": 4.889645768054483e-06,
-      "loss": 1.7052,
+      "loss": 1.7817,
       "step": 5200
     },
     {
       "epoch": 0.106,
-      "grad_norm": 2.3854475021362305,
+      "grad_norm": 2.762117624282837,
       "learning_rate": 4.884935611950279e-06,
-      "loss": 1.6863,
+      "loss": 1.7917,
       "step": 5300
     },
     {
       "epoch": 0.108,
-      "grad_norm": 2.598034620285034,
+      "grad_norm": 2.7164549827575684,
       "learning_rate": 4.8801293909873806e-06,
-      "loss": 1.7042,
+      "loss": 1.7807,
       "step": 5400
     },
     {
       "epoch": 0.11,
-      "grad_norm": 2.382126569747925,
+      "grad_norm": 2.7724523544311523,
       "learning_rate": 4.8752272987596675e-06,
-      "loss": 1.6838,
+      "loss": 1.7624,
       "step": 5500
     },
     {
       "epoch": 0.112,
-      "grad_norm": 2.7612802982330322,
+      "grad_norm": 2.5202503204345703,
       "learning_rate": 4.870229532722699e-06,
-      "loss": 1.7027,
+      "loss": 1.7646,
       "step": 5600
     },
     {
       "epoch": 0.114,
-      "grad_norm": 2.736612558364868,
+      "grad_norm": 2.5876247882843018,
       "learning_rate": 4.865136294185763e-06,
-      "loss": 1.7243,
+      "loss": 1.7488,
       "step": 5700
     },
     {
       "epoch": 0.116,
-      "grad_norm": 2.7250194549560547,
+      "grad_norm": 2.848844051361084,
       "learning_rate": 4.859947788303764e-06,
-      "loss": 1.7021,
+      "loss": 1.7623,
       "step": 5800
     },
     {
       "epoch": 0.118,
-      "grad_norm": 2.718186140060425,
+      "grad_norm": 2.5332720279693604,
       "learning_rate": 4.854664224068964e-06,
-      "loss": 1.6849,
+      "loss": 1.7404,
       "step": 5900
     },
     {
       "epoch": 0.12,
-      "grad_norm": 2.501079797744751,
+      "grad_norm": 2.4839518070220947,
       "learning_rate": 4.8492858143025575e-06,
-      "loss": 1.7058,
+      "loss": 1.7487,
       "step": 6000
     },
     {
       "epoch": 0.12,
-      "eval_loss": 1.676255464553833,
-      "eval_runtime": 24.1172,
-      "eval_samples_per_second": 207.321,
-      "eval_steps_per_second": 3.276,
+      "eval_loss": 1.7817753553390503,
+      "eval_runtime": 23.8863,
+      "eval_samples_per_second": 209.325,
+      "eval_steps_per_second": 3.307,
       "step": 6000
     },
     {
       "epoch": 0.122,
-      "grad_norm": 2.4820473194122314,
+      "grad_norm": 2.717193841934204,
       "learning_rate": 4.8438127756461095e-06,
-      "loss": 1.7046,
+      "loss": 1.751,
       "step": 6100
     },
     {
       "epoch": 0.124,
-      "grad_norm": 2.5755321979522705,
+      "grad_norm": 2.454843044281006,
       "learning_rate": 4.838245328552815e-06,
-      "loss": 1.6998,
+      "loss": 1.7574,
       "step": 6200
     },
     {
       "epoch": 0.126,
-      "grad_norm": 2.4646663665771484,
+      "grad_norm": 2.5411481857299805,
       "learning_rate": 4.832583697278635e-06,
-      "loss": 1.6996,
+      "loss": 1.7518,
       "step": 6300
     },
     {
       "epoch": 0.128,
-      "grad_norm": 2.6041085720062256,
+      "grad_norm": 2.4669787883758545,
       "learning_rate": 4.826828109873248e-06,
-      "loss": 1.6993,
+      "loss": 1.7489,
       "step": 6400
     },
     {
       "epoch": 0.13,
-      "grad_norm": 2.793649196624756,
+      "grad_norm": 2.4951584339141846,
       "learning_rate": 4.8209787981708775e-06,
-      "loss": 1.715,
+      "loss": 1.7768,
       "step": 6500
     },
     {
       "epoch": 0.132,
-      "grad_norm": 2.4691555500030518,
+      "grad_norm": 2.824913263320923,
       "learning_rate": 4.815035997780945e-06,
-      "loss": 1.6824,
+      "loss": 1.7512,
       "step": 6600
     },
     {
       "epoch": 0.134,
-      "grad_norm": 2.404930353164673,
+      "grad_norm": 2.48067307472229,
       "learning_rate": 4.8089999480785854e-06,
-      "loss": 1.7016,
+      "loss": 1.7627,
       "step": 6700
     },
     {
       "epoch": 0.136,
-      "grad_norm": 2.665576219558716,
+      "grad_norm": 2.7985410690307617,
       "learning_rate": 4.802870892194999e-06,
-      "loss": 1.6835,
+      "loss": 1.7466,
       "step": 6800
     },
     {
       "epoch": 0.138,
-      "grad_norm": 2.521707773208618,
+      "grad_norm": 2.6190848350524902,
       "learning_rate": 4.796649077007663e-06,
-      "loss": 1.6851,
+      "loss": 1.7546,
       "step": 6900
     },
     {
       "epoch": 0.14,
-      "grad_norm": 2.3326563835144043,
+      "grad_norm": 2.7269248962402344,
       "learning_rate": 4.790334753130387e-06,
-      "loss": 1.6926,
+      "loss": 1.7482,
       "step": 7000
     },
     {
       "epoch": 0.14,
-      "eval_loss": 1.711079716682434,
-      "eval_runtime": 24.1632,
-      "eval_samples_per_second": 206.927,
-      "eval_steps_per_second": 3.269,
+      "eval_loss": 1.7849754095077515,
+      "eval_runtime": 23.8694,
+      "eval_samples_per_second": 209.473,
+      "eval_steps_per_second": 3.31,
       "step": 7000
     },
     {
       "epoch": 0.142,
-      "grad_norm": 2.782231569290161,
+      "grad_norm": 2.703486680984497,
       "learning_rate": 4.783928174903215e-06,
-      "loss": 1.6856,
+      "loss": 1.7488,
       "step": 7100
     },
     {
       "epoch": 0.144,
-      "grad_norm": 2.5902581214904785,
+      "grad_norm": 2.7523727416992188,
       "learning_rate": 4.7774296003821855e-06,
-      "loss": 1.6978,
+      "loss": 1.7492,
       "step": 7200
     },
     {
       "epoch": 0.146,
-      "grad_norm": 2.7924344539642334,
+      "grad_norm": 2.6879053115844727,
       "learning_rate": 4.77083929132893e-06,
-      "loss": 1.686,
+      "loss": 1.7577,
       "step": 7300
     },
     {
       "epoch": 0.148,
-      "grad_norm": 2.3744988441467285,
+      "grad_norm": 2.5821115970611572,
       "learning_rate": 4.764157513200138e-06,
-      "loss": 1.6988,
+      "loss": 1.7529,
       "step": 7400
     },
     {
       "epoch": 0.15,
-      "grad_norm": 2.309220552444458,
+      "grad_norm": 2.569995880126953,
       "learning_rate": 4.757384535136858e-06,
-      "loss": 1.6969,
+      "loss": 1.7284,
       "step": 7500
     },
     {
       "epoch": 0.152,
-      "grad_norm": 2.708876609802246,
+      "grad_norm": 2.7729454040527344,
       "learning_rate": 4.750520629953659e-06,
-      "loss": 1.7011,
+      "loss": 1.7506,
       "step": 7600
     },
     {
       "epoch": 0.154,
-      "grad_norm": 2.4460690021514893,
+      "grad_norm": 2.629626989364624,
       "learning_rate": 4.74356607412764e-06,
-      "loss": 1.7056,
+      "loss": 1.7655,
       "step": 7700
     },
     {
       "epoch": 0.156,
-      "grad_norm": 2.602306842803955,
+      "grad_norm": 2.409274101257324,
       "learning_rate": 4.736521147787297e-06,
-      "loss": 1.6886,
+      "loss": 1.7492,
       "step": 7800
     },
     {
       "epoch": 0.158,
-      "grad_norm": 2.689602851867676,
+      "grad_norm": 2.537017583847046,
       "learning_rate": 4.729386134701235e-06,
-      "loss": 1.6895,
+      "loss": 1.7623,
       "step": 7900
     },
     {
       "epoch": 0.16,
-      "grad_norm": 2.5177204608917236,
+      "grad_norm": 2.92161226272583,
       "learning_rate": 4.7221613222667395e-06,
-      "loss": 1.6929,
+      "loss": 1.7566,
       "step": 8000
     },
     {
       "epoch": 0.16,
-      "eval_loss": 1.683371663093567,
-      "eval_runtime": 24.1907,
-      "eval_samples_per_second": 206.691,
-      "eval_steps_per_second": 3.266,
+      "eval_loss": 1.7861406803131104,
+      "eval_runtime": 23.9237,
+      "eval_samples_per_second": 208.998,
+      "eval_steps_per_second": 3.302,
       "step": 8000
     },
     {
       "epoch": 0.162,
-      "grad_norm": 2.903104543685913,
+      "grad_norm": 2.6894986629486084,
       "learning_rate": 4.714847001498204e-06,
-      "loss": 1.7375,
+      "loss": 1.7348,
       "step": 8100
     },
     {
       "epoch": 0.164,
-      "grad_norm": 2.7971339225769043,
+      "grad_norm": 2.725722551345825,
       "learning_rate": 4.707443467015401e-06,
-      "loss": 1.7943,
+      "loss": 1.7266,
       "step": 8200
     },
     {
       "epoch": 0.166,
-      "grad_norm": 2.700718641281128,
+      "grad_norm": 2.3508708477020264,
       "learning_rate": 4.699951017031622e-06,
-      "loss": 1.7982,
+      "loss": 1.7598,
       "step": 8300
     },
     {
       "epoch": 0.168,
-      "grad_norm": 2.8577218055725098,
+      "grad_norm": 2.7327499389648438,
       "learning_rate": 4.692369953341657e-06,
-      "loss": 1.7941,
+      "loss": 1.7409,
       "step": 8400
     },
     {
       "epoch": 0.17,
-      "grad_norm": 2.4928951263427734,
+      "grad_norm": 2.745523691177368,
       "learning_rate": 4.684700581309648e-06,
-      "loss": 1.8113,
+      "loss": 1.7473,
       "step": 8500
     },
     {
       "epoch": 0.172,
-      "grad_norm": 2.9383528232574463,
+      "grad_norm": 2.646702289581299,
       "learning_rate": 4.676943209856781e-06,
-      "loss": 1.7867,
+      "loss": 1.7584,
       "step": 8600
     },
     {
       "epoch": 0.174,
-      "grad_norm": 2.79278302192688,
+      "grad_norm": 2.797407388687134,
       "learning_rate": 4.669098151448848e-06,
-      "loss": 1.8073,
+      "loss": 1.7566,
       "step": 8700
     },
     {
       "epoch": 0.176,
-      "grad_norm": 2.751046895980835,
+      "grad_norm": 2.845431089401245,
       "learning_rate": 4.661165722083658e-06,
-      "loss": 1.7908,
+      "loss": 1.7359,
       "step": 8800
     },
     {
       "epoch": 0.178,
-      "grad_norm": 2.775073766708374,
+      "grad_norm": 2.7239723205566406,
       "learning_rate": 4.653146241278308e-06,
-      "loss": 1.7807,
+      "loss": 1.7523,
       "step": 8900
     },
     {
       "epoch": 0.18,
-      "grad_norm": 2.946188449859619,
+      "grad_norm": 2.532681465148926,
       "learning_rate": 4.645040032056316e-06,
-      "loss": 1.7817,
+      "loss": 1.7403,
       "step": 9000
     },
     {
       "epoch": 0.18,
-      "eval_loss": 1.694687843322754,
-      "eval_runtime": 24.2446,
-      "eval_samples_per_second": 206.232,
-      "eval_steps_per_second": 3.258,
+      "eval_loss": 1.7844035625457764,
+      "eval_runtime": 23.9573,
+      "eval_samples_per_second": 208.704,
+      "eval_steps_per_second": 3.298,
       "step": 9000
     },
     {
       "epoch": 0.182,
-      "grad_norm": 2.9802608489990234,
+      "grad_norm": 2.5148870944976807,
       "learning_rate": 4.636847420934607e-06,
-      "loss": 1.8123,
+      "loss": 1.7543,
       "step": 9100
     },
     {
       "epoch": 0.184,
-      "grad_norm": 2.5945792198181152,
+      "grad_norm": 2.6170496940612793,
       "learning_rate": 4.6285687379103635e-06,
-      "loss": 1.7849,
+      "loss": 1.7366,
       "step": 9200
     },
     {
       "epoch": 0.186,
-      "grad_norm": 2.737100601196289,
+      "grad_norm": 2.7940258979797363,
       "learning_rate": 4.620204316447729e-06,
-      "loss": 1.7924,
+      "loss": 1.7157,
       "step": 9300
     },
     {
       "epoch": 0.188,
-      "grad_norm": 3.0725672245025635,
+      "grad_norm": 2.7664384841918945,
       "learning_rate": 4.611754493464383e-06,
-      "loss": 1.8107,
+      "loss": 1.7306,
       "step": 9400
     },
     {
       "epoch": 0.19,
-      "grad_norm": 2.5555238723754883,
+      "grad_norm": 2.4932405948638916,
       "learning_rate": 4.603219609317961e-06,
-      "loss": 1.7949,
+      "loss": 1.7591,
       "step": 9500
     },
     {
       "epoch": 0.192,
-      "grad_norm": 2.5154004096984863,
+      "grad_norm": 2.7944176197052,
       "learning_rate": 4.594600007792352e-06,
-      "loss": 1.7863,
+      "loss": 1.7507,
       "step": 9600
     },
     {
       "epoch": 0.194,
-      "grad_norm": 2.891353130340576,
+      "grad_norm": 2.6389007568359375,
       "learning_rate": 4.585896036083849e-06,
-      "loss": 1.7922,
+      "loss": 1.7425,
       "step": 9700
     },
     {
       "epoch": 0.196,
-      "grad_norm": 2.678168773651123,
+      "grad_norm": 2.577260971069336,
       "learning_rate": 4.577108044787163e-06,
-      "loss": 1.8023,
+      "loss": 1.755,
       "step": 9800
     },
     {
       "epoch": 0.198,
-      "grad_norm": 2.7289631366729736,
+      "grad_norm": 2.3340439796447754,
       "learning_rate": 4.568236387881303e-06,
-      "loss": 1.805,
+      "loss": 1.7566,
       "step": 9900
     },
     {
       "epoch": 0.2,
-      "grad_norm": 2.695540189743042,
+      "grad_norm": 2.7538001537323,
       "learning_rate": 4.5592814227153145e-06,
-      "loss": 1.8079,
+      "loss": 1.7576,
       "step": 10000
     },
     {
       "epoch": 0.2,
-      "eval_loss": 1.6816699504852295,
-      "eval_runtime": 24.2484,
-      "eval_samples_per_second": 206.199,
-      "eval_steps_per_second": 3.258,
+      "eval_loss": 1.7423070669174194,
+      "eval_runtime": 23.9875,
+      "eval_samples_per_second": 208.442,
+      "eval_steps_per_second": 3.293,
       "step": 10000
     },
     {
       "epoch": 0.202,
-      "grad_norm": 2.6634812355041504,
+      "grad_norm": 2.58754825592041,
       "learning_rate": 4.550243509993889e-06,
-      "loss": 1.7831,
+      "loss": 1.7373,
       "step": 10100
     },
     {
       "epoch": 0.204,
-      "grad_norm": 2.622002363204956,
+      "grad_norm": 2.478538990020752,
       "learning_rate": 4.541123013762833e-06,
-      "loss": 1.7894,
+      "loss": 1.7243,
       "step": 10200
     },
     {
       "epoch": 0.206,
-      "grad_norm": 2.689647674560547,
+      "grad_norm": 2.665117025375366,
       "learning_rate": 4.531920301394405e-06,
-      "loss": 1.7893,
+      "loss": 1.7301,
       "step": 10300
     },
     {
       "epoch": 0.208,
-      "grad_norm": 2.6407201290130615,
+      "grad_norm": 2.7941126823425293,
       "learning_rate": 4.5226357435725175e-06,
-      "loss": 1.8027,
+      "loss": 1.7515,
       "step": 10400
     },
     {
       "epoch": 0.21,
-      "grad_norm": 2.5345051288604736,
+      "grad_norm": 2.446150302886963,
       "learning_rate": 4.5132697142778045e-06,
-      "loss": 1.8043,
+      "loss": 1.7417,
       "step": 10500
     },
     {
       "epoch": 0.212,
-      "grad_norm": 2.658268690109253,
+      "grad_norm": 2.705165386199951,
       "learning_rate": 4.503822590772563e-06,
-      "loss": 1.7857,
+      "loss": 1.7261,
       "step": 10600
     },
     {
       "epoch": 0.214,
-      "grad_norm": 2.663511276245117,
+      "grad_norm": 2.8230743408203125,
       "learning_rate": 4.494294753585547e-06,
-      "loss": 1.789,
+      "loss": 1.7355,
       "step": 10700
     },
     {
       "epoch": 0.216,
-      "grad_norm": 2.973478317260742,
+      "grad_norm": 2.519519090652466,
       "learning_rate": 4.484686586496652e-06,
-      "loss": 1.7893,
+      "loss": 1.7325,
       "step": 10800
     },
     {
       "epoch": 0.218,
-      "grad_norm": 2.588331937789917,
+      "grad_norm": 2.53291392326355,
       "learning_rate": 4.474998476521445e-06,
-      "loss": 1.8135,
+      "loss": 1.7372,
       "step": 10900
     },
     {
       "epoch": 0.22,
-      "grad_norm": 2.800774574279785,
+      "grad_norm": 2.6000454425811768,
       "learning_rate": 4.465230813895585e-06,
-      "loss": 1.7879,
+      "loss": 1.7537,
       "step": 11000
     },
     {
       "epoch": 0.22,
-      "eval_loss": 1.7125898599624634,
-      "eval_runtime": 24.2117,
-      "eval_samples_per_second": 206.511,
-      "eval_steps_per_second": 3.263,
+      "eval_loss": 1.7882028818130493,
+      "eval_runtime": 24.0299,
+      "eval_samples_per_second": 208.074,
+      "eval_steps_per_second": 3.288,
       "step": 11000
     },
     {
       "epoch": 0.222,
-      "grad_norm": 2.7460155487060547,
+      "grad_norm": 2.837104082107544,
       "learning_rate": 4.455383992059097e-06,
-      "loss": 1.7877,
+      "loss": 1.7282,
       "step": 11100
     },
     {
       "epoch": 0.224,
-      "grad_norm": 2.6012282371520996,
+      "grad_norm": 2.6326394081115723,
       "learning_rate": 4.4454584076405285e-06,
-      "loss": 1.7931,
+      "loss": 1.7339,
       "step": 11200
     },
     {
       "epoch": 0.226,
-      "grad_norm": 2.9110138416290283,
+      "grad_norm": 2.670193910598755,
       "learning_rate": 4.435454460440974e-06,
-      "loss": 1.797,
+      "loss": 1.7455,
       "step": 11300
     },
     {
       "epoch": 0.228,
-      "grad_norm": 2.7014622688293457,
+      "grad_norm": 2.5847089290618896,
       "learning_rate": 4.425372553417968e-06,
-      "loss": 1.7827,
+      "loss": 1.7337,
       "step": 11400
     },
     {
       "epoch": 0.23,
-      "grad_norm": 2.582334518432617,
+      "grad_norm": 2.5647053718566895,
       "learning_rate": 4.415213092669254e-06,
-      "loss": 1.7804,
+      "loss": 1.745,
       "step": 11500
     },
     {
       "epoch": 0.232,
-      "grad_norm": 2.565603494644165,
+      "grad_norm": 2.3335916996002197,
       "learning_rate": 4.40497648741643e-06,
-      "loss": 1.7826,
+      "loss": 1.75,
       "step": 11600
     },
     {
       "epoch": 0.234,
-      "grad_norm": 2.8357231616973877,
+      "grad_norm": 2.696769952774048,
       "learning_rate": 4.394663149988464e-06,
-      "loss": 1.8059,
+      "loss": 1.7612,
       "step": 11700
     },
     {
       "epoch": 0.236,
-      "grad_norm": 2.5272321701049805,
+      "grad_norm": 2.4234607219696045,
       "learning_rate": 4.3842734958050826e-06,
-      "loss": 1.7858,
+      "loss": 1.7119,
       "step": 11800
     },
     {
       "epoch": 0.238,
-      "grad_norm": 2.772449493408203,
+      "grad_norm": 2.9543232917785645,
       "learning_rate": 4.373807943360041e-06,
-      "loss": 1.7779,
+      "loss": 1.7135,
       "step": 11900
     },
     {
       "epoch": 0.24,
-      "grad_norm": 2.4692089557647705,
+      "grad_norm": 2.7549548149108887,
       "learning_rate": 4.363266914204267e-06,
-      "loss": 1.8047,
+      "loss": 1.7342,
       "step": 12000
     },
     {
       "epoch": 0.24,
-      "eval_loss": 1.706162929534912,
-      "eval_runtime": 24.3218,
-      "eval_samples_per_second": 205.577,
-      "eval_steps_per_second": 3.248,
+      "eval_loss": 1.7919249534606934,
+      "eval_runtime": 24.0703,
+      "eval_samples_per_second": 207.725,
+      "eval_steps_per_second": 3.282,
       "step": 12000
     },
     {
       "epoch": 0.242,
-      "grad_norm": 2.5936381816864014,
+      "grad_norm": 2.7592766284942627,
       "learning_rate": 4.352650832928878e-06,
-      "loss": 1.7841,
+      "loss": 1.7405,
       "step": 12100
     },
     {
       "epoch": 0.244,
-      "grad_norm": 2.65234637260437,
+      "grad_norm": 2.5189287662506104,
       "learning_rate": 4.341960127148078e-06,
-      "loss": 1.7908,
+      "loss": 1.7426,
       "step": 12200
     },
     {
       "epoch": 0.246,
-      "grad_norm": 2.763434886932373,
+      "grad_norm": 2.576111078262329,
       "learning_rate": 4.331195227481939e-06,
-      "loss": 1.7964,
+      "loss": 1.7246,
       "step": 12300
     },
     {
       "epoch": 0.248,
-      "grad_norm": 2.785336971282959,
+      "grad_norm": 2.866229772567749,
       "learning_rate": 4.320356567539049e-06,
-      "loss": 1.8119,
+      "loss": 1.7275,
       "step": 12400
     },
     {
       "epoch": 0.25,
-      "grad_norm": 2.8722548484802246,
+      "grad_norm": 2.4703452587127686,
       "learning_rate": 4.30944458389905e-06,
-      "loss": 1.7921,
+      "loss": 1.7336,
       "step": 12500
     },
     {
       "epoch": 0.252,
-      "grad_norm": 2.5671935081481934,
+      "grad_norm": 2.812617540359497,
       "learning_rate": 4.298459716095053e-06,
-      "loss": 1.7801,
+      "loss": 1.7454,
       "step": 12600
     },
     {
       "epoch": 0.254,
-      "grad_norm": 2.761291027069092,
+      "grad_norm": 2.586317777633667,
       "learning_rate": 4.2874024065959295e-06,
-      "loss": 1.7877,
+      "loss": 1.7392,
       "step": 12700
     },
     {
       "epoch": 0.256,
-      "grad_norm": 2.638906240463257,
+      "grad_norm": 2.85270357131958,
       "learning_rate": 4.276273100788495e-06,
-      "loss": 1.7723,
+      "loss": 1.7368,
       "step": 12800
     },
     {
       "epoch": 0.258,
-      "grad_norm": 2.706116199493408,
+      "grad_norm": 2.690568447113037,
       "learning_rate": 4.265072246959564e-06,
-      "loss": 1.7822,
+      "loss": 1.7222,
       "step": 12900
     },
     {
       "epoch": 0.26,
-      "grad_norm": 2.706327199935913,
+      "grad_norm": 2.9626104831695557,
       "learning_rate": 4.253800296277897e-06,
-      "loss": 1.7932,
+      "loss": 1.7301,
       "step": 13000
     },
     {
       "epoch": 0.26,
-      "eval_loss": 1.6767820119857788,
-      "eval_runtime": 24.2948,
-      "eval_samples_per_second": 205.806,
-      "eval_steps_per_second": 3.252,
+      "eval_loss": 1.7579104900360107,
+      "eval_runtime": 24.0451,
+      "eval_samples_per_second": 207.942,
+      "eval_steps_per_second": 3.285,
       "step": 13000
     },
     {
       "epoch": 0.262,
-      "grad_norm": 2.532775640487671,
+      "grad_norm": 2.629026174545288,
       "learning_rate": 4.242457702776022e-06,
-      "loss": 1.7707,
+      "loss": 1.7291,
       "step": 13100
     },
     {
       "epoch": 0.264,
-      "grad_norm": 3.2018957138061523,
+      "grad_norm": 2.970782995223999,
       "learning_rate": 4.2310449233319525e-06,
-      "loss": 1.7924,
+      "loss": 1.7348,
       "step": 13200
     },
     {
       "epoch": 0.266,
-      "grad_norm": 2.7994580268859863,
+      "grad_norm": 2.6203627586364746,
       "learning_rate": 4.21956241765078e-06,
-      "loss": 1.775,
+      "loss": 1.748,
       "step": 13300
     },
     {
       "epoch": 0.268,
-      "grad_norm": 2.8110687732696533,
+      "grad_norm": 2.931784152984619,
       "learning_rate": 4.208010648246158e-06,
-      "loss": 1.7793,
+      "loss": 1.7412,
       "step": 13400
     },
     {
       "epoch": 0.27,
-      "grad_norm": 2.786374092102051,
+      "grad_norm": 2.657593250274658,
       "learning_rate": 4.1963900804216765e-06,
-      "loss": 1.7954,
+      "loss": 1.7294,
       "step": 13500
     },
     {
       "epoch": 0.272,
-      "grad_norm": 2.586460828781128,
+      "grad_norm": 2.6762797832489014,
       "learning_rate": 4.18470118225211e-06,
-      "loss": 1.7688,
+      "loss": 1.7299,
       "step": 13600
     },
     {
       "epoch": 0.274,
-      "grad_norm": 2.813493490219116,
+      "grad_norm": 2.760854482650757,
       "learning_rate": 4.172944424564574e-06,
-      "loss": 1.7803,
+      "loss": 1.7434,
       "step": 13700
     },
     {
       "epoch": 0.276,
-      "grad_norm": 2.752819538116455,
+      "grad_norm": 2.6275932788848877,
       "learning_rate": 4.161120280919555e-06,
-      "loss": 1.8007,
+      "loss": 1.7209,
       "step": 13800
     },
     {
       "epoch": 0.278,
-      "grad_norm": 2.7272469997406006,
+      "grad_norm": 2.7294981479644775,
       "learning_rate": 4.1492292275918315e-06,
-      "loss": 1.7729,
+      "loss": 1.7186,
       "step": 13900
     },
     {
       "epoch": 0.28,
-      "grad_norm": 2.820791721343994,
+      "grad_norm": 2.5626657009124756,
       "learning_rate": 4.1372717435513006e-06,
-      "loss": 1.7936,
+      "loss": 1.7429,
       "step": 14000
     },
     {
       "epoch": 0.28,
-      "eval_loss": 1.71174156665802,
-      "eval_runtime": 24.4957,
-      "eval_samples_per_second": 204.118,
-      "eval_steps_per_second": 3.225,
+      "eval_loss": 1.7644519805908203,
+      "eval_runtime": 24.1148,
+      "eval_samples_per_second": 207.341,
+      "eval_steps_per_second": 3.276,
       "step": 14000
     },
     {
       "epoch": 0.282,
-      "grad_norm": 2.699657440185547,
+      "grad_norm": 2.386528968811035,
       "learning_rate": 4.125248310443675e-06,
-      "loss": 1.7846,
+      "loss": 1.7334,
       "step": 14100
     },
     {
       "epoch": 0.284,
-      "grad_norm": 2.7167606353759766,
+      "grad_norm": 2.9119138717651367,
       "learning_rate": 4.113159412571087e-06,
-      "loss": 1.7792,
+      "loss": 1.7305,
       "step": 14200
     },
     {
       "epoch": 0.286,
-      "grad_norm": 2.615206241607666,
+      "grad_norm": 2.513732671737671,
       "learning_rate": 4.10100553687258e-06,
-      "loss": 1.7665,
+      "loss": 1.7377,
       "step": 14300
     },
     {
       "epoch": 0.288,
-      "grad_norm": 2.646576166152954,
+      "grad_norm": 2.599806547164917,
       "learning_rate": 4.088787172904496e-06,
-      "loss": 1.7995,
+      "loss": 1.728,
       "step": 14400
     },
     {
       "epoch": 0.29,
-      "grad_norm": 2.8086984157562256,
+      "grad_norm": 2.561784029006958,
       "learning_rate": 4.076504812820752e-06,
-      "loss": 1.787,
+      "loss": 1.7191,
       "step": 14500
     },
     {
       "epoch": 0.292,
-      "grad_norm": 2.868511915206909,
+      "grad_norm": 2.601694345474243,
       "learning_rate": 4.064158951353024e-06,
-      "loss": 1.7957,
+      "loss": 1.7248,
       "step": 14600
     },
     {
       "epoch": 0.294,
-      "grad_norm": 2.696380376815796,
+      "grad_norm": 2.74916410446167,
       "learning_rate": 4.051750085790809e-06,
-      "loss": 1.7916,
+      "loss": 1.7563,
       "step": 14700
     },
     {
       "epoch": 0.296,
-      "grad_norm": 2.7678110599517822,
+      "grad_norm": 2.7767183780670166,
       "learning_rate": 4.039278715961403e-06,
-      "loss": 1.7927,
+      "loss": 1.7257,
       "step": 14800
     },
     {
       "epoch": 0.298,
-      "grad_norm": 2.54803729057312,
+      "grad_norm": 2.529806613922119,
       "learning_rate": 4.026745344209766e-06,
-      "loss": 1.7792,
+      "loss": 1.7209,
       "step": 14900
     },
     {
       "epoch": 0.3,
-      "grad_norm": 2.63761568069458,
+      "grad_norm": 2.4386274814605713,
       "learning_rate": 4.014150475378284e-06,
-      "loss": 1.7764,
+      "loss": 1.7362,
       "step": 15000
     },
     {
       "epoch": 0.3,
-      "eval_loss": 1.7043291330337524,
-      "eval_runtime": 24.4661,
-      "eval_samples_per_second": 204.365,
-      "eval_steps_per_second": 3.229,
+      "eval_loss": 1.773182988166809,
+      "eval_runtime": 24.0902,
+      "eval_samples_per_second": 207.553,
+      "eval_steps_per_second": 3.279,
       "step": 15000
     },
     {
       "epoch": 0.302,
-      "grad_norm": 2.774980306625366,
+      "grad_norm": 2.486804485321045,
       "learning_rate": 4.001494616786436e-06,
-      "loss": 1.7557,
+      "loss": 1.7402,
       "step": 15100
     },
     {
       "epoch": 0.304,
-      "grad_norm": 2.5435757637023926,
+      "grad_norm": 2.438204050064087,
       "learning_rate": 3.9887782782103615e-06,
-      "loss": 1.7676,
+      "loss": 1.7295,
       "step": 15200
     },
     {
       "epoch": 0.306,
-      "grad_norm": 2.513993740081787,
+      "grad_norm": 2.552276372909546,
       "learning_rate": 3.976001971862323e-06,
-      "loss": 1.775,
+      "loss": 1.7203,
       "step": 15300
     },
     {
       "epoch": 0.308,
-      "grad_norm": 3.040370464324951,
+      "grad_norm": 2.8073322772979736,
       "learning_rate": 3.963166212370078e-06,
-      "loss": 1.7797,
+      "loss": 1.7279,
       "step": 15400
     },
     {
       "epoch": 0.31,
-      "grad_norm": 2.7786099910736084,
+      "grad_norm": 2.575695753097534,
       "learning_rate": 3.9502715167561486e-06,
-      "loss": 1.7761,
+      "loss": 1.7466,
       "step": 15500
     },
     {
       "epoch": 0.312,
-      "grad_norm": 2.717729330062866,
+      "grad_norm": 2.7184011936187744,
       "learning_rate": 3.937318404416993e-06,
-      "loss": 1.7469,
+      "loss": 1.7223,
       "step": 15600
     },
     {
       "epoch": 0.314,
-      "grad_norm": 2.4638280868530273,
+      "grad_norm": 2.660555839538574,
       "learning_rate": 3.924307397102086e-06,
-      "loss": 1.7775,
+      "loss": 1.7352,
       "step": 15700
     },
     {
       "epoch": 0.316,
-      "grad_norm": 2.6738526821136475,
+      "grad_norm": 2.3733389377593994,
       "learning_rate": 3.911239018892907e-06,
-      "loss": 1.7866,
+      "loss": 1.74,
       "step": 15800
     },
     {
       "epoch": 0.318,
-      "grad_norm": 2.587191581726074,
+      "grad_norm": 2.596536874771118,
       "learning_rate": 3.898113796181824e-06,
-      "loss": 1.7693,
+      "loss": 1.7084,
       "step": 15900
     },
     {
       "epoch": 0.32,
-      "grad_norm": 2.81728196144104,
+      "grad_norm": 2.5756585597991943,
       "learning_rate": 3.8849322576508934e-06,
-      "loss": 1.7836,
+      "loss": 1.7444,
       "step": 16000
     },
     {
       "epoch": 0.32,
-      "eval_loss": 1.6930394172668457,
-      "eval_runtime": 24.353,
-      "eval_samples_per_second": 205.314,
-      "eval_steps_per_second": 3.244,
+      "eval_loss": 1.7792890071868896,
+      "eval_runtime": 24.1638,
+      "eval_samples_per_second": 206.921,
+      "eval_steps_per_second": 3.269,
       "step": 16000
     },
     {
       "epoch": 0.322,
-      "grad_norm": 2.668858289718628,
+      "grad_norm": 2.748867988586426,
       "learning_rate": 3.871694934250566e-06,
-      "loss": 1.7754,
+      "loss": 1.7392,
       "step": 16100
     },
     {
       "epoch": 0.324,
-      "grad_norm": 2.631056308746338,
+      "grad_norm": 2.5147626399993896,
       "learning_rate": 3.858402359178298e-06,
-      "loss": 1.7747,
+      "loss": 1.7443,
       "step": 16200
     },
     {
       "epoch": 0.326,
-      "grad_norm": 2.7213263511657715,
+      "grad_norm": 2.686070680618286,
       "learning_rate": 3.845055067857075e-06,
-      "loss": 1.7862,
+      "loss": 1.743,
       "step": 16300
     },
     {
       "epoch": 0.328,
-      "grad_norm": 2.6167287826538086,
+      "grad_norm": 2.737027645111084,
       "learning_rate": 3.8316535979138454e-06,
-      "loss": 1.7927,
+      "loss": 1.7425,
       "step": 16400
     },
     {
       "epoch": 0.33,
-      "grad_norm": 2.46602201461792,
+      "grad_norm": 2.7026660442352295,
       "learning_rate": 3.818198489157864e-06,
-      "loss": 1.7958,
+      "loss": 1.7082,
       "step": 16500
     },
     {
       "epoch": 0.332,
-      "grad_norm": 2.773110866546631,
+      "grad_norm": 2.431051015853882,
       "learning_rate": 3.8046902835589505e-06,
-      "loss": 1.7691,
+      "loss": 1.7163,
       "step": 16600
     },
     {
       "epoch": 0.334,
-      "grad_norm": 2.4469850063323975,
+      "grad_norm": 2.46547269821167,
       "learning_rate": 3.7911295252256554e-06,
-      "loss": 1.7799,
+      "loss": 1.748,
       "step": 16700
     },
     {
       "epoch": 0.336,
-      "grad_norm": 2.667153835296631,
+      "grad_norm": 2.805881977081299,
       "learning_rate": 3.777516760383347e-06,
-      "loss": 1.7829,
+      "loss": 1.7198,
       "step": 16800
     },
     {
       "epoch": 0.338,
-      "grad_norm": 2.5449864864349365,
+      "grad_norm": 2.628716230392456,
       "learning_rate": 3.7638525373522096e-06,
-      "loss": 1.7832,
+      "loss": 1.7413,
       "step": 16900
     },
     {
       "epoch": 0.34,
-      "grad_norm": 2.9086837768554688,
+      "grad_norm": 2.4161248207092285,
       "learning_rate": 3.750137406525154e-06,
-      "loss": 1.7723,
+      "loss": 1.7285,
       "step": 17000
     },
     {
       "epoch": 0.34,
-      "eval_loss": 1.6615562438964844,
-      "eval_runtime": 25.324,
-      "eval_samples_per_second": 197.441,
-      "eval_steps_per_second": 3.12,
+      "eval_loss": 1.7723710536956787,
+      "eval_runtime": 24.3183,
+      "eval_samples_per_second": 205.607,
+      "eval_steps_per_second": 3.249,
       "step": 17000
     },
     {
       "epoch": 0.342,
-      "grad_norm": 2.889113187789917,
+      "grad_norm": 2.442286491394043,
       "learning_rate": 3.7363719203456493e-06,
-      "loss": 1.7966,
+      "loss": 1.736,
       "step": 17100
     },
     {
       "epoch": 0.344,
-      "grad_norm": 2.667297840118408,
+      "grad_norm": 2.559293270111084,
       "learning_rate": 3.7225566332854724e-06,
-      "loss": 1.7758,
+      "loss": 1.7304,
       "step": 17200
     },
     {
       "epoch": 0.346,
-      "grad_norm": 2.8233277797698975,
+      "grad_norm": 2.4408977031707764,
       "learning_rate": 3.7086921018223722e-06,
-      "loss": 1.7911,
+      "loss": 1.7115,
       "step": 17300
     },
     {
       "epoch": 0.348,
-      "grad_norm": 2.9642388820648193,
+      "grad_norm": 2.6720967292785645,
       "learning_rate": 3.6947788844176523e-06,
-      "loss": 1.7727,
+      "loss": 1.733,
       "step": 17400
     },
     {
       "epoch": 0.35,
-      "grad_norm": 2.7055301666259766,
+      "grad_norm": 2.8161537647247314,
       "learning_rate": 3.680817541493683e-06,
-      "loss": 1.7857,
+      "loss": 1.7315,
       "step": 17500
     },
     {
       "epoch": 0.352,
-      "grad_norm": 2.6071488857269287,
+      "grad_norm": 2.348341226577759,
       "learning_rate": 3.6668086354113208e-06,
-      "loss": 1.7839,
+      "loss": 1.713,
       "step": 17600
     },
     {
       "epoch": 0.354,
-      "grad_norm": 2.963099479675293,
+      "grad_norm": 2.8460476398468018,
       "learning_rate": 3.6527527304472603e-06,
-      "loss": 1.7811,
+      "loss": 1.7348,
       "step": 17700
     },
     {
       "epoch": 0.356,
-      "grad_norm": 2.7885355949401855,
+      "grad_norm": 3.1378066539764404,
       "learning_rate": 3.6386503927713036e-06,
-      "loss": 1.783,
+      "loss": 1.716,
       "step": 17800
     },
     {
       "epoch": 0.358,
-      "grad_norm": 2.5152931213378906,
+      "grad_norm": 2.5782840251922607,
       "learning_rate": 3.624502190423556e-06,
-      "loss": 1.7882,
+      "loss": 1.7285,
       "step": 17900
     },
     {
       "epoch": 0.36,
-      "grad_norm": 2.799872875213623,
+      "grad_norm": 2.5901360511779785,
       "learning_rate": 3.6103086932915454e-06,
-      "loss": 1.7756,
+      "loss": 1.7253,
       "step": 18000
     },
     {
       "epoch": 0.36,
-      "eval_loss": 1.7140281200408936,
-      "eval_runtime": 24.4132,
-      "eval_samples_per_second": 204.807,
-      "eval_steps_per_second": 3.236,
+      "eval_loss": 1.7683390378952026,
+      "eval_runtime": 25.2647,
+      "eval_samples_per_second": 197.905,
+      "eval_steps_per_second": 3.127,
       "step": 18000
     },
     {
       "epoch": 0.362,
-      "grad_norm": 2.9310367107391357,
+      "grad_norm": 2.582977771759033,
       "learning_rate": 3.5960704730872657e-06,
-      "loss": 1.7712,
+      "loss": 1.7264,
       "step": 18100
     },
     {
       "epoch": 0.364,
-      "grad_norm": 2.79758620262146,
+      "grad_norm": 2.6602444648742676,
       "learning_rate": 3.5817881033241518e-06,
-      "loss": 1.7596,
+      "loss": 1.723,
       "step": 18200
     },
     {
       "epoch": 0.366,
-      "grad_norm": 2.688558578491211,
+      "grad_norm": 2.931091070175171,
       "learning_rate": 3.5674621592939753e-06,
-      "loss": 1.764,
+      "loss": 1.7333,
       "step": 18300
     },
     {
       "epoch": 0.368,
-      "grad_norm": 2.473888397216797,
+      "grad_norm": 2.880857229232788,
       "learning_rate": 3.5530932180436723e-06,
-      "loss": 1.7747,
+      "loss": 1.727,
       "step": 18400
     },
     {
       "epoch": 0.37,
-      "grad_norm": 2.9678609371185303,
+      "grad_norm": 2.631927967071533,
       "learning_rate": 3.5386818583521003e-06,
-      "loss": 1.7674,
+      "loss": 1.7328,
       "step": 18500
     },
     {
       "epoch": 0.372,
-      "grad_norm": 2.6792540550231934,
+      "grad_norm": 2.7482428550720215,
       "learning_rate": 3.5242286607067277e-06,
-      "loss": 1.7772,
+      "loss": 1.7347,
       "step": 18600
     },
     {
       "epoch": 0.374,
-      "grad_norm": 2.3820743560791016,
+      "grad_norm": 2.8233249187469482,
       "learning_rate": 3.509734207280247e-06,
-      "loss": 1.744,
+      "loss": 1.7194,
       "step": 18700
     },
     {
       "epoch": 0.376,
-      "grad_norm": 2.578467607498169,
+      "grad_norm": 2.6914608478546143,
       "learning_rate": 3.495199081907131e-06,
-      "loss": 1.7793,
+      "loss": 1.7357,
       "step": 18800
     },
     {
       "epoch": 0.378,
-      "grad_norm": 2.5736358165740967,
+      "grad_norm": 2.548243999481201,
       "learning_rate": 3.4806238700601073e-06,
-      "loss": 1.769,
+      "loss": 1.733,
       "step": 18900
     },
     {
       "epoch": 0.38,
-      "grad_norm": 2.5953547954559326,
+      "grad_norm": 2.7307281494140625,
       "learning_rate": 3.466009158826586e-06,
-      "loss": 1.7821,
+      "loss": 1.7309,
       "step": 19000
     },
     {
       "epoch": 0.38,
-      "eval_loss": 1.6795705556869507,
-      "eval_runtime": 24.3829,
-      "eval_samples_per_second": 205.061,
-      "eval_steps_per_second": 3.24,
+      "eval_loss": 1.7916011810302734,
+      "eval_runtime": 24.2592,
+      "eval_samples_per_second": 206.108,
+      "eval_steps_per_second": 3.256,
       "step": 19000
     },
     {
       "epoch": 0.382,
-      "grad_norm": 2.9146714210510254,
+      "grad_norm": 2.9709601402282715,
       "learning_rate": 3.451355536885004e-06,
-      "loss": 1.7684,
+      "loss": 1.709,
       "step": 19100
     },
     {
       "epoch": 0.384,
-      "grad_norm": 2.6636972427368164,
+      "grad_norm": 2.8460004329681396,
       "learning_rate": 3.4366635944811155e-06,
-      "loss": 1.7871,
+      "loss": 1.7386,
       "step": 19200
     },
     {
       "epoch": 0.386,
-      "grad_norm": 2.9465370178222656,
+      "grad_norm": 2.907198429107666,
       "learning_rate": 3.421933923404219e-06,
-      "loss": 1.7822,
+      "loss": 1.7226,
       "step": 19300
     },
     {
       "epoch": 0.388,
-      "grad_norm": 2.8429131507873535,
+      "grad_norm": 2.747058391571045,
       "learning_rate": 3.4071671169633186e-06,
-      "loss": 1.7743,
+      "loss": 1.7081,
       "step": 19400
     },
     {
       "epoch": 0.39,
-      "grad_norm": 2.571214437484741,
+      "grad_norm": 3.0197839736938477,
       "learning_rate": 3.3923637699632235e-06,
-      "loss": 1.7829,
+      "loss": 1.7438,
       "step": 19500
     },
     {
       "epoch": 0.392,
-      "grad_norm": 2.740962028503418,
+      "grad_norm": 2.5856029987335205,
       "learning_rate": 3.377524478680592e-06,
-      "loss": 1.7723,
+      "loss": 1.7225,
       "step": 19600
     },
     {
       "epoch": 0.394,
-      "grad_norm": 2.7149062156677246,
+      "grad_norm": 2.782837390899658,
       "learning_rate": 3.3626498408399142e-06,
-      "loss": 1.767,
+      "loss": 1.7497,
       "step": 19700
     },
     {
       "epoch": 0.396,
-      "grad_norm": 2.6767420768737793,
+      "grad_norm": 2.638728618621826,
       "learning_rate": 3.347740455589432e-06,
-      "loss": 1.7931,
+      "loss": 1.7277,
       "step": 19800
     },
     {
       "epoch": 0.398,
-      "grad_norm": 2.8027760982513428,
+      "grad_norm": 2.800845146179199,
       "learning_rate": 3.3327969234770117e-06,
-      "loss": 1.7606,
+      "loss": 1.7332,
       "step": 19900
     },
     {
       "epoch": 0.4,
-      "grad_norm": 2.917006492614746,
+      "grad_norm": 2.582824945449829,
       "learning_rate": 3.3178198464259454e-06,
-      "loss": 1.7865,
+      "loss": 1.7273,
       "step": 20000
     },
     {
       "epoch": 0.4,
-      "eval_loss": 1.6951992511749268,
-      "eval_runtime": 24.4103,
-      "eval_samples_per_second": 204.832,
-      "eval_steps_per_second": 3.236,
+      "eval_loss": 1.7508941888809204,
+      "eval_runtime": 24.3003,
+      "eval_samples_per_second": 205.759,
+      "eval_steps_per_second": 3.251,
       "step": 20000
     },
     {
       "epoch": 0.402,
-      "grad_norm": 2.6517817974090576,
+      "grad_norm": 2.538066864013672,
       "learning_rate": 3.302809827710714e-06,
-      "loss": 1.7735,
+      "loss": 1.6906,
       "step": 20100
     },
     {
       "epoch": 0.404,
-      "grad_norm": 2.6432833671569824,
+      "grad_norm": 2.5869054794311523,
       "learning_rate": 3.2877674719326823e-06,
-      "loss": 1.7743,
+      "loss": 1.7127,
       "step": 20200
     },
     {
       "epoch": 0.406,
-      "grad_norm": 2.566873073577881,
+      "grad_norm": 2.474055767059326,
       "learning_rate": 3.272693384995749e-06,
-      "loss": 1.7642,
+      "loss": 1.727,
       "step": 20300
     },
     {
       "epoch": 0.408,
-      "grad_norm": 2.6067020893096924,
+      "grad_norm": 2.6444203853607178,
       "learning_rate": 3.2575881740819352e-06,
-      "loss": 1.7601,
+      "loss": 1.7346,
       "step": 20400
     },
     {
       "epoch": 0.41,
-      "grad_norm": 2.5541176795959473,
+      "grad_norm": 2.61991286277771,
       "learning_rate": 3.2424524476269364e-06,
-      "loss": 1.7625,
+      "loss": 1.7194,
       "step": 20500
     },
     {
       "epoch": 0.412,
-      "grad_norm": 2.9584436416625977,
+      "grad_norm": 2.64259934425354,
       "learning_rate": 3.227286815295606e-06,
-      "loss": 1.7739,
+      "loss": 1.7185,
       "step": 20600
     },
     {
       "epoch": 0.414,
-      "grad_norm": 2.6747519969940186,
+      "grad_norm": 2.608701705932617,
       "learning_rate": 3.2120918879574043e-06,
-      "loss": 1.7834,
+      "loss": 1.7292,
       "step": 20700
     },
     {
       "epoch": 0.416,
-      "grad_norm": 2.6932618618011475,
+      "grad_norm": 2.5797171592712402,
       "learning_rate": 3.196868277661787e-06,
-      "loss": 1.7676,
+      "loss": 1.7024,
       "step": 20800
     },
     {
       "epoch": 0.418,
-      "grad_norm": 2.5855135917663574,
+      "grad_norm": 2.5299649238586426,
       "learning_rate": 3.181616597613558e-06,
-      "loss": 1.7764,
+      "loss": 1.7176,
       "step": 20900
     },
     {
       "epoch": 0.42,
-      "grad_norm": 2.595672607421875,
+      "grad_norm": 2.541083812713623,
       "learning_rate": 3.166337462148166e-06,
-      "loss": 1.7858,
+      "loss": 1.7194,
       "step": 21000
     },
     {
       "epoch": 0.42,
-      "eval_loss": 1.6869325637817383,
-      "eval_runtime": 24.4337,
-      "eval_samples_per_second": 204.635,
-      "eval_steps_per_second": 3.233,
+      "eval_loss": 1.770919919013977,
+      "eval_runtime": 24.2944,
+      "eval_samples_per_second": 205.809,
+      "eval_steps_per_second": 3.252,
       "step": 21000
     },
     {
       "epoch": 0.422,
-      "grad_norm": 2.760918378829956,
+      "grad_norm": 2.859060525894165,
       "learning_rate": 3.1510314867069596e-06,
-      "loss": 1.7644,
+      "loss": 1.7296,
       "step": 21100
     },
     {
       "epoch": 0.424,
-      "grad_norm": 2.762570381164551,
+      "grad_norm": 2.659062623977661,
       "learning_rate": 3.135699287812397e-06,
-      "loss": 1.7624,
+      "loss": 1.7018,
       "step": 21200
     },
     {
       "epoch": 0.426,
-      "grad_norm": 2.771238088607788,
+      "grad_norm": 2.4112436771392822,
       "learning_rate": 3.120341483043215e-06,
-      "loss": 1.7598,
+      "loss": 1.7202,
       "step": 21300
     },
     {
       "epoch": 0.428,
-      "grad_norm": 2.8537368774414062,
+      "grad_norm": 2.3485844135284424,
       "learning_rate": 3.104958691009549e-06,
-      "loss": 1.7765,
+      "loss": 1.7678,
       "step": 21400
     },
     {
       "epoch": 0.43,
-      "grad_norm": 2.7945148944854736,
+      "grad_norm": 1.9426896572113037,
       "learning_rate": 3.089551531328021e-06,
-      "loss": 1.7779,
+      "loss": 1.7741,
       "step": 21500
     },
     {
       "epoch": 0.432,
-      "grad_norm": 2.8764634132385254,
+      "grad_norm": 1.8812894821166992,
       "learning_rate": 3.0741206245967747e-06,
-      "loss": 1.7661,
+      "loss": 1.7612,
       "step": 21600
     },
     {
       "epoch": 0.434,
-      "grad_norm": 2.593735933303833,
+      "grad_norm": 2.720904588699341,
       "learning_rate": 3.0586665923704837e-06,
-      "loss": 1.7796,
+      "loss": 1.7782,
       "step": 21700
     },
     {
       "epoch": 0.436,
-      "grad_norm": 2.5224249362945557,
+      "grad_norm": 2.078925132751465,
       "learning_rate": 3.043190057135314e-06,
-      "loss": 1.7914,
+      "loss": 1.7608,
       "step": 21800
     },
     {
       "epoch": 0.438,
-      "grad_norm": 2.587533712387085,
+      "grad_norm": 2.197183609008789,
       "learning_rate": 3.0276916422838465e-06,
-      "loss": 1.7709,
+      "loss": 1.7833,
       "step": 21900
     },
     {
       "epoch": 0.44,
-      "grad_norm": 2.942674160003662,
+      "grad_norm": 1.9943153858184814,
       "learning_rate": 3.0121719720899715e-06,
-      "loss": 1.763,
+      "loss": 1.7595,
       "step": 22000
     },
     {
       "epoch": 0.44,
-      "eval_loss": 1.6977226734161377,
-      "eval_runtime": 24.4615,
-      "eval_samples_per_second": 204.403,
-      "eval_steps_per_second": 3.23,
+      "eval_loss": 1.767109751701355,
+      "eval_runtime": 24.3398,
+      "eval_samples_per_second": 205.425,
+      "eval_steps_per_second": 3.246,
       "step": 22000
     },
     {
       "epoch": 0.442,
-      "grad_norm": 2.7402122020721436,
+      "grad_norm": 1.9623926877975464,
       "learning_rate": 2.9966316716837424e-06,
-      "loss": 1.7998,
+      "loss": 1.76,
       "step": 22100
     },
     {
       "epoch": 0.444,
-      "grad_norm": 2.5526087284088135,
+      "grad_norm": 1.900605320930481,
       "learning_rate": 2.9810713670261938e-06,
-      "loss": 1.7612,
+      "loss": 1.7577,
       "step": 22200
     },
     {
       "epoch": 0.446,
-      "grad_norm": 2.795562744140625,
+      "grad_norm": 1.9600718021392822,
       "learning_rate": 2.9654916848841276e-06,
-      "loss": 1.7748,
+      "loss": 1.7556,
       "step": 22300
     },
     {
       "epoch": 0.448,
-      "grad_norm": 2.783095121383667,
+      "grad_norm": 1.9493147134780884,
       "learning_rate": 2.9498932528048685e-06,
-      "loss": 1.7611,
+      "loss": 1.7639,
       "step": 22400
     },
     {
       "epoch": 0.45,
-      "grad_norm": 2.581273317337036,
+      "grad_norm": 2.3909194469451904,
       "learning_rate": 2.9342766990909865e-06,
-      "loss": 1.7671,
+      "loss": 1.7672,
       "step": 22500
     },
     {
       "epoch": 0.452,
-      "grad_norm": 2.6591572761535645,
+      "grad_norm": 2.3083951473236084,
       "learning_rate": 2.918642652774989e-06,
-      "loss": 1.7651,
+      "loss": 1.7555,
       "step": 22600
     },
     {
       "epoch": 0.454,
-      "grad_norm": 2.6311802864074707,
+      "grad_norm": 1.8799562454223633,
       "learning_rate": 2.9029917435939812e-06,
-      "loss": 1.7698,
+      "loss": 1.774,
       "step": 22700
     },
     {
       "epoch": 0.456,
-      "grad_norm": 2.6259610652923584,
+      "grad_norm": 2.2513976097106934,
       "learning_rate": 2.8873246019643026e-06,
-      "loss": 1.7601,
+      "loss": 1.7552,
       "step": 22800
     },
     {
       "epoch": 0.458,
-      "grad_norm": 2.9224014282226562,
+      "grad_norm": 2.050868511199951,
       "learning_rate": 2.871641858956134e-06,
-      "loss": 1.7685,
+      "loss": 1.7555,
       "step": 22900
     },
     {
       "epoch": 0.46,
-      "grad_norm": 2.794389247894287,
+      "grad_norm": 2.1203055381774902,
       "learning_rate": 2.855944146268077e-06,
-      "loss": 1.7764,
+      "loss": 1.7526,
       "step": 23000
     },
     {
       "epoch": 0.46,
-      "eval_loss": 1.6730164289474487,
-      "eval_runtime": 24.4416,
-      "eval_samples_per_second": 204.569,
-      "eval_steps_per_second": 3.232,
+      "eval_loss": 1.7840089797973633,
+      "eval_runtime": 24.3706,
+      "eval_samples_per_second": 205.165,
+      "eval_steps_per_second": 3.242,
       "step": 23000
     },
     {
       "epoch": 0.462,
-      "grad_norm": 2.680636167526245,
+      "grad_norm": 1.9163652658462524,
       "learning_rate": 2.8402320962017094e-06,
-      "loss": 1.7636,
+      "loss": 1.7522,
       "step": 23100
     },
     {
       "epoch": 0.464,
-      "grad_norm": 2.6686863899230957,
+      "grad_norm": 2.2142231464385986,
       "learning_rate": 2.824506341636117e-06,
-      "loss": 1.768,
+      "loss": 1.7508,
       "step": 23200
     },
     {
       "epoch": 0.466,
-      "grad_norm": 2.8378450870513916,
+      "grad_norm": 1.969416856765747,
       "learning_rate": 2.808767516002401e-06,
-      "loss": 1.765,
+      "loss": 1.7471,
       "step": 23300
     },
     {
       "epoch": 0.468,
-      "grad_norm": 2.7039575576782227,
+      "grad_norm": 1.7630442380905151,
       "learning_rate": 2.7930162532581623e-06,
-      "loss": 1.781,
+      "loss": 1.7872,
       "step": 23400
     },
     {
       "epoch": 0.47,
-      "grad_norm": 2.370654582977295,
+      "grad_norm": 1.765560269355774,
       "learning_rate": 2.7772531878619685e-06,
-      "loss": 1.7702,
+      "loss": 1.7475,
       "step": 23500
     },
     {
       "epoch": 0.472,
-      "grad_norm": 2.6227610111236572,
+      "grad_norm": 1.8466675281524658,
       "learning_rate": 2.7614789547477944e-06,
-      "loss": 1.7615,
+      "loss": 1.7486,
       "step": 23600
     },
     {
       "epoch": 0.474,
-      "grad_norm": 2.704012870788574,
+      "grad_norm": 2.2333123683929443,
       "learning_rate": 2.7456941892994497e-06,
-      "loss": 1.7642,
+      "loss": 1.7408,
       "step": 23700
     },
     {
       "epoch": 0.476,
-      "grad_norm": 2.7952730655670166,
+      "grad_norm": 1.9206023216247559,
       "learning_rate": 2.729899527324986e-06,
-      "loss": 1.7645,
+      "loss": 1.7573,
       "step": 23800
     },
     {
       "epoch": 0.478,
-      "grad_norm": 2.732647657394409,
+      "grad_norm": 1.8804105520248413,
       "learning_rate": 2.714095605031084e-06,
-      "loss": 1.7683,
+      "loss": 1.7448,
       "step": 23900
     },
     {
       "epoch": 0.48,
-      "grad_norm": 2.5511538982391357,
+      "grad_norm": 2.1470730304718018,
       "learning_rate": 2.698283058997429e-06,
-      "loss": 1.7778,
+      "loss": 1.7425,
       "step": 24000
     },
     {
       "epoch": 0.48,
-      "eval_loss": 1.6797676086425781,
-      "eval_runtime": 24.4385,
-      "eval_samples_per_second": 204.596,
-      "eval_steps_per_second": 3.233,
+      "eval_loss": 1.7723808288574219,
+      "eval_runtime": 24.3905,
+      "eval_samples_per_second": 204.998,
+      "eval_steps_per_second": 3.239,
       "step": 24000
     },
     {
       "epoch": 0.482,
-      "grad_norm": 2.7857868671417236,
+      "grad_norm": 1.9350930452346802,
       "learning_rate": 2.6824625261510695e-06,
-      "loss": 1.7631,
+      "loss": 1.7655,
       "step": 24100
     },
     {
       "epoch": 0.484,
-      "grad_norm": 2.796555757522583,
+      "grad_norm": 1.94557785987854,
       "learning_rate": 2.666634643740763e-06,
-      "loss": 1.7699,
+      "loss": 1.7523,
       "step": 24200
     },
     {
       "epoch": 0.486,
-      "grad_norm": 2.763692855834961,
+      "grad_norm": 2.1901168823242188,
       "learning_rate": 2.650800049311304e-06,
-      "loss": 1.7554,
+      "loss": 1.7535,
       "step": 24300
     },
     {
       "epoch": 0.488,
-      "grad_norm": 2.743865966796875,
+      "grad_norm": 2.181553840637207,
       "learning_rate": 2.634959380677847e-06,
-      "loss": 1.763,
+      "loss": 1.7421,
       "step": 24400
     },
     {
       "epoch": 0.49,
-      "grad_norm": 2.7378954887390137,
+      "grad_norm": 1.872516393661499,
       "learning_rate": 2.6191132759002148e-06,
-      "loss": 1.7735,
+      "loss": 1.7428,
       "step": 24500
     },
     {
       "epoch": 0.492,
-      "grad_norm": 2.554868221282959,
+      "grad_norm": 2.0588748455047607,
       "learning_rate": 2.6032623732571976e-06,
-      "loss": 1.7901,
+      "loss": 1.7643,
       "step": 24600
     },
     {
       "epoch": 0.494,
-      "grad_norm": 2.9881274700164795,
+      "grad_norm": 1.8576792478561401,
       "learning_rate": 2.587407311220843e-06,
-      "loss": 1.7736,
+      "loss": 1.7555,
       "step": 24700
     },
     {
       "epoch": 0.496,
-      "grad_norm": 2.818943500518799,
+      "grad_norm": 1.9538050889968872,
       "learning_rate": 2.571548728430737e-06,
-      "loss": 1.7644,
+      "loss": 1.7422,
       "step": 24800
     },
     {
       "epoch": 0.498,
-      "grad_norm": 2.6200406551361084,
+      "grad_norm": 2.1427853107452393,
       "learning_rate": 2.555687263668284e-06,
-      "loss": 1.781,
+      "loss": 1.7512,
       "step": 24900
     },
     {
       "epoch": 0.5,
-      "grad_norm": 2.7953360080718994,
+      "grad_norm": 1.9427967071533203,
       "learning_rate": 2.5398235558309688e-06,
-      "loss": 1.7927,
+      "loss": 1.7342,
       "step": 25000
     },
     {
       "epoch": 0.5,
-      "eval_loss": 1.6672734022140503,
-      "eval_runtime": 24.6709,
-      "eval_samples_per_second": 202.668,
-      "eval_steps_per_second": 3.202,
+      "eval_loss": 1.775884985923767,
+      "eval_runtime": 24.366,
+      "eval_samples_per_second": 205.204,
+      "eval_steps_per_second": 3.242,
       "step": 25000
     },
     {
       "epoch": 0.502,
-      "grad_norm": 2.8738749027252197,
+      "grad_norm": 2.0338659286499023,
       "learning_rate": 2.5239582439066316e-06,
-      "loss": 1.838,
+      "loss": 1.7455,
       "step": 25100
     },
     {
       "epoch": 0.504,
-      "grad_norm": 2.779383659362793,
+      "grad_norm": 1.9146922826766968,
       "learning_rate": 2.508091966947724e-06,
-      "loss": 1.8433,
+      "loss": 1.7351,
       "step": 25200
     },
     {
       "epoch": 0.506,
-      "grad_norm": 3.024062156677246,
+      "grad_norm": 1.827868103981018,
       "learning_rate": 2.4922253640455676e-06,
-      "loss": 1.8631,
+      "loss": 1.7559,
       "step": 25300
     },
     {
       "epoch": 0.508,
-      "grad_norm": 2.728076219558716,
+      "grad_norm": 2.316199541091919,
       "learning_rate": 2.4763590743046135e-06,
-      "loss": 1.8562,
+      "loss": 1.7523,
       "step": 25400
     },
     {
       "epoch": 0.51,
-      "grad_norm": 2.6469523906707764,
+      "grad_norm": 1.9336920976638794,
       "learning_rate": 2.460493736816699e-06,
-      "loss": 1.8753,
+      "loss": 1.745,
       "step": 25500
     },
     {
       "epoch": 0.512,
-      "grad_norm": 3.0614748001098633,
+      "grad_norm": 1.9351345300674438,
       "learning_rate": 2.444629990635305e-06,
-      "loss": 1.8779,
+      "loss": 1.7286,
       "step": 25600
     },
     {
       "epoch": 0.514,
-      "grad_norm": 4.503206729888916,
+      "grad_norm": 2.255042791366577,
       "learning_rate": 2.4287684747498153e-06,
-      "loss": 1.852,
+      "loss": 1.7577,
       "step": 25700
     },
     {
       "epoch": 0.516,
-      "grad_norm": 3.175844430923462,
+      "grad_norm": 2.241882562637329,
       "learning_rate": 2.4129098280597767e-06,
-      "loss": 1.8349,
+      "loss": 1.7407,
       "step": 25800
     },
     {
       "epoch": 0.518,
-      "grad_norm": 3.0593061447143555,
+      "grad_norm": 2.121722936630249,
       "learning_rate": 2.3970546893491637e-06,
-      "loss": 1.8487,
+      "loss": 1.7484,
       "step": 25900
     },
     {
       "epoch": 0.52,
-      "grad_norm": 2.934358596801758,
+      "grad_norm": 1.9267939329147339,
       "learning_rate": 2.3812036972606527e-06,
-      "loss": 1.8528,
+      "loss": 1.7438,
       "step": 26000
     },
     {
       "epoch": 0.52,
-      "eval_loss": 1.6837133169174194,
-      "eval_runtime": 24.6678,
-      "eval_samples_per_second": 202.694,
-      "eval_steps_per_second": 3.203,
+      "eval_loss": 1.778982400894165,
+      "eval_runtime": 24.3959,
+      "eval_samples_per_second": 204.953,
+      "eval_steps_per_second": 3.238,
       "step": 26000
     },
     {
       "epoch": 0.522,
-      "grad_norm": 2.577605724334717,
+      "grad_norm": 2.1775739192962646,
       "learning_rate": 2.365357490269895e-06,
-      "loss": 1.8596,
+      "loss": 1.73,
       "step": 26100
     },
     {
       "epoch": 0.524,
-      "grad_norm": 2.7109344005584717,
+      "grad_norm": 1.9757606983184814,
       "learning_rate": 2.349516706659796e-06,
-      "loss": 1.8511,
+      "loss": 1.7567,
       "step": 26200
     },
     {
       "epoch": 0.526,
-      "grad_norm": 2.8878512382507324,
+      "grad_norm": 1.8726664781570435,
       "learning_rate": 2.3336819844948105e-06,
-      "loss": 1.856,
+      "loss": 1.7275,
       "step": 26300
     },
     {
       "epoch": 0.528,
-      "grad_norm": 2.831289052963257,
+      "grad_norm": 1.9961349964141846,
       "learning_rate": 2.3178539615952385e-06,
-      "loss": 1.8483,
+      "loss": 1.7453,
       "step": 26400
     },
     {
       "epoch": 0.53,
-      "grad_norm": 2.664224863052368,
+      "grad_norm": 2.1781928539276123,
       "learning_rate": 2.3020332755115336e-06,
-      "loss": 1.8697,
+      "loss": 1.7386,
       "step": 26500
     },
     {
       "epoch": 0.532,
-      "grad_norm": 2.67335844039917,
+      "grad_norm": 1.9620516300201416,
       "learning_rate": 2.286220563498625e-06,
-      "loss": 1.8595,
+      "loss": 1.7336,
       "step": 26600
     },
     {
       "epoch": 0.534,
-      "grad_norm": 2.6854031085968018,
+      "grad_norm": 1.786254644393921,
       "learning_rate": 2.270416462490248e-06,
-      "loss": 1.8372,
+      "loss": 1.7447,
       "step": 26700
     },
     {
       "epoch": 0.536,
-      "grad_norm": 2.6764156818389893,
+      "grad_norm": 1.8988018035888672,
       "learning_rate": 2.2546216090732844e-06,
-      "loss": 1.8366,
+      "loss": 1.7179,
       "step": 26800
     },
     {
       "epoch": 0.538,
-      "grad_norm": 2.534574270248413,
+      "grad_norm": 2.260758638381958,
       "learning_rate": 2.2388366394621286e-06,
-      "loss": 1.8314,
+      "loss": 1.7439,
       "step": 26900
     },
     {
       "epoch": 0.54,
-      "grad_norm": 2.9713919162750244,
+      "grad_norm": 2.1915199756622314,
       "learning_rate": 2.223062189473054e-06,
-      "loss": 1.8281,
+      "loss": 1.7304,
       "step": 27000
     },
     {
       "epoch": 0.54,
-      "eval_loss": 1.6805105209350586,
-      "eval_runtime": 24.4746,
-      "eval_samples_per_second": 204.293,
-      "eval_steps_per_second": 3.228,
+      "eval_loss": 1.7679800987243652,
+      "eval_runtime": 24.3919,
+      "eval_samples_per_second": 204.986,
+      "eval_steps_per_second": 3.239,
       "step": 27000
     },
     {
       "epoch": 0.542,
-      "grad_norm": 2.9666595458984375,
+      "grad_norm": 1.9719111919403076,
       "learning_rate": 2.2072988944986056e-06,
-      "loss": 1.8404,
+      "loss": 1.7446,
       "step": 27100
     },
     {
       "epoch": 0.544,
-      "grad_norm": 2.7935214042663574,
+      "grad_norm": 1.9965158700942993,
       "learning_rate": 2.1915473894820083e-06,
-      "loss": 1.8289,
+      "loss": 1.7252,
       "step": 27200
     },
     {
       "epoch": 0.546,
-      "grad_norm": 2.8710145950317383,
+      "grad_norm": 1.9306226968765259,
       "learning_rate": 2.1758083088915852e-06,
-      "loss": 1.84,
+      "loss": 1.7438,
       "step": 27300
     },
     {
       "epoch": 0.548,
-      "grad_norm": 2.7820184230804443,
+      "grad_norm": 2.143664598464966,
       "learning_rate": 2.1600822866952093e-06,
-      "loss": 1.8387,
+      "loss": 1.7363,
       "step": 27400
     },
     {
       "epoch": 0.55,
-      "grad_norm": 2.768270492553711,
+      "grad_norm": 1.9158095121383667,
       "learning_rate": 2.144369956334762e-06,
-      "loss": 1.8334,
+      "loss": 1.7326,
       "step": 27500
     },
     {
       "epoch": 0.552,
-      "grad_norm": 2.595914125442505,
+      "grad_norm": 1.9001954793930054,
       "learning_rate": 2.1286719507006163e-06,
-      "loss": 1.8378,
+      "loss": 1.7215,
       "step": 27600
     },
     {
       "epoch": 0.554,
-      "grad_norm": 2.8370561599731445,
+      "grad_norm": 1.883360505104065,
       "learning_rate": 2.112988902106153e-06,
-      "loss": 1.8292,
+      "loss": 1.734,
       "step": 27700
     },
     {
       "epoch": 0.556,
-      "grad_norm": 2.807756185531616,
+      "grad_norm": 2.2805025577545166,
       "learning_rate": 2.0973214422622807e-06,
-      "loss": 1.8456,
+      "loss": 1.7346,
       "step": 27800
     },
     {
       "epoch": 0.558,
-      "grad_norm": 2.8122119903564453,
+      "grad_norm": 1.7587010860443115,
       "learning_rate": 2.0816702022519983e-06,
-      "loss": 1.8355,
+      "loss": 1.7529,
       "step": 27900
     },
     {
       "epoch": 0.56,
-      "grad_norm": 2.731086015701294,
+      "grad_norm": 1.7039998769760132,
       "learning_rate": 2.066035812504972e-06,
-      "loss": 1.8346,
+      "loss": 1.7153,
       "step": 28000
     },
     {
       "epoch": 0.56,
-      "eval_loss": 1.6786452531814575,
-      "eval_runtime": 24.4812,
-      "eval_samples_per_second": 204.239,
-      "eval_steps_per_second": 3.227,
+      "eval_loss": 1.7555021047592163,
+      "eval_runtime": 24.4164,
+      "eval_samples_per_second": 204.781,
+      "eval_steps_per_second": 3.236,
       "step": 28000
     },
     {
       "epoch": 0.562,
-      "grad_norm": 2.8046417236328125,
+      "grad_norm": 1.9466190338134766,
       "learning_rate": 2.0504189027721396e-06,
-      "loss": 1.8373,
+      "loss": 1.7344,
       "step": 28100
     },
     {
       "epoch": 0.564,
-      "grad_norm": 3.015007495880127,
+      "grad_norm": 1.9495512247085571,
       "learning_rate": 2.034820102100349e-06,
-      "loss": 1.8474,
+      "loss": 1.7354,
       "step": 28200
     },
     {
       "epoch": 0.566,
-      "grad_norm": 2.770026206970215,
+      "grad_norm": 1.8734641075134277,
       "learning_rate": 2.0192400388070173e-06,
-      "loss": 1.8472,
+      "loss": 1.7416,
       "step": 28300
     },
     {
       "epoch": 0.568,
-      "grad_norm": 2.856353759765625,
+      "grad_norm": 1.92928946018219,
       "learning_rate": 2.003679340454821e-06,
-      "loss": 1.8313,
+      "loss": 1.7388,
       "step": 28400
     },
     {
       "epoch": 0.57,
-      "grad_norm": 2.6885221004486084,
+      "grad_norm": 1.9698309898376465,
       "learning_rate": 1.988138633826421e-06,
-      "loss": 1.8398,
+      "loss": 1.7239,
       "step": 28500
     },
     {
       "epoch": 0.572,
-      "grad_norm": 2.9842429161071777,
+      "grad_norm": 2.2002367973327637,
       "learning_rate": 1.9726185448992143e-06,
-      "loss": 1.836,
+      "loss": 1.7333,
       "step": 28600
     },
     {
       "epoch": 0.574,
-      "grad_norm": 2.8554325103759766,
+      "grad_norm": 1.968303918838501,
       "learning_rate": 1.95711969882012e-06,
-      "loss": 1.8211,
+      "loss": 1.7296,
       "step": 28700
     },
     {
       "epoch": 0.576,
-      "grad_norm": 2.6220102310180664,
+      "grad_norm": 1.9326261281967163,
       "learning_rate": 1.9416427198803987e-06,
-      "loss": 1.8234,
+      "loss": 1.708,
       "step": 28800
     },
     {
       "epoch": 0.578,
-      "grad_norm": 2.66099214553833,
+      "grad_norm": 1.995526909828186,
       "learning_rate": 1.926188231490504e-06,
-      "loss": 1.8354,
+      "loss": 1.7364,
       "step": 28900
     },
     {
       "epoch": 0.58,
-      "grad_norm": 2.6300864219665527,
+      "grad_norm": 1.6052130460739136,
       "learning_rate": 1.910756856154976e-06,
-      "loss": 1.8298,
+      "loss": 1.7528,
       "step": 29000
     },
     {
       "epoch": 0.58,
-      "eval_loss": 1.6784433126449585,
-      "eval_runtime": 24.4774,
-      "eval_samples_per_second": 204.27,
+      "eval_loss": 1.7665057182312012,
+      "eval_runtime": 24.481,
+      "eval_samples_per_second": 204.24,
       "eval_steps_per_second": 3.227,
       "step": 29000
     },
     {
       "epoch": 0.582,
-      "grad_norm": 2.7466800212860107,
+      "grad_norm": 1.0827213525772095,
       "learning_rate": 1.8953492154473634e-06,
-      "loss": 1.8394,
+      "loss": 1.7529,
       "step": 29100
     },
     {
       "epoch": 0.584,
-      "grad_norm": 2.6594157218933105,
+      "grad_norm": 0.959848940372467,
       "learning_rate": 1.879965929985187e-06,
-      "loss": 1.8274,
+      "loss": 1.7543,
       "step": 29200
     },
     {
       "epoch": 0.586,
-      "grad_norm": 2.8458616733551025,
+      "grad_norm": 0.9542915225028992,
       "learning_rate": 1.8646076194049405e-06,
-      "loss": 1.8367,
+      "loss": 1.7612,
       "step": 29300
     },
     {
       "epoch": 0.588,
-      "grad_norm": 2.5822947025299072,
+      "grad_norm": 0.9279881119728088,
       "learning_rate": 1.8492749023371365e-06,
-      "loss": 1.8292,
+      "loss": 1.7674,
       "step": 29400
     },
     {
       "epoch": 0.59,
-      "grad_norm": 2.872380495071411,
+      "grad_norm": 1.001535177230835,
       "learning_rate": 1.8339683963813815e-06,
-      "loss": 1.829,
+      "loss": 1.7528,
       "step": 29500
     },
     {
       "epoch": 0.592,
-      "grad_norm": 2.673823118209839,
+      "grad_norm": 0.9729927182197571,
       "learning_rate": 1.8186887180815051e-06,
-      "loss": 1.8406,
+      "loss": 1.7472,
       "step": 29600
     },
     {
       "epoch": 0.594,
-      "grad_norm": 2.8034603595733643,
+      "grad_norm": 0.945178210735321,
       "learning_rate": 1.8034364829007198e-06,
-      "loss": 1.8481,
+      "loss": 1.7524,
       "step": 29700
     },
     {
       "epoch": 0.596,
-      "grad_norm": 2.6873667240142822,
+      "grad_norm": 1.0035436153411865,
       "learning_rate": 1.7882123051968354e-06,
-      "loss": 1.8242,
+      "loss": 1.7423,
       "step": 29800
     },
     {
       "epoch": 0.598,
-      "grad_norm": 2.707566261291504,
+      "grad_norm": 0.9502407908439636,
       "learning_rate": 1.77301679819751e-06,
-      "loss": 1.8339,
+      "loss": 1.7418,
       "step": 29900
     },
     {
       "epoch": 0.6,
-      "grad_norm": 2.8388428688049316,
+      "grad_norm": 0.9792412519454956,
       "learning_rate": 1.7578505739755491e-06,
-      "loss": 1.8372,
+      "loss": 1.7419,
       "step": 30000
     },
     {
       "epoch": 0.6,
-      "eval_loss": 1.6919031143188477,
-      "eval_runtime": 25.4878,
-      "eval_samples_per_second": 196.172,
-      "eval_steps_per_second": 3.1,
+      "eval_loss": 1.7764304876327515,
+      "eval_runtime": 24.5814,
+      "eval_samples_per_second": 203.406,
+      "eval_steps_per_second": 3.214,
       "step": 30000
     },
     {
       "epoch": 0.602,
-      "grad_norm": 2.5644257068634033,
+      "grad_norm": 0.9637987017631531,
       "learning_rate": 1.7427142434242544e-06,
-      "loss": 1.8249,
+      "loss": 1.7433,
       "step": 30100
     },
     {
       "epoch": 0.604,
-      "grad_norm": 2.868516683578491,
+      "grad_norm": 0.9773419499397278,
       "learning_rate": 1.7276084162328117e-06,
-      "loss": 1.8352,
+      "loss": 1.743,
       "step": 30200
     },
     {
       "epoch": 0.606,
-      "grad_norm": 2.8583498001098633,
+      "grad_norm": 0.9079156517982483,
       "learning_rate": 1.7125337008617387e-06,
-      "loss": 1.8303,
+      "loss": 1.7357,
       "step": 30300
     },
     {
       "epoch": 0.608,
-      "grad_norm": 2.896786689758301,
+      "grad_norm": 0.9946913123130798,
       "learning_rate": 1.6974907045183719e-06,
-      "loss": 1.8509,
+      "loss": 1.7373,
       "step": 30400
     },
     {
       "epoch": 0.61,
-      "grad_norm": 2.864170789718628,
+      "grad_norm": 0.9330785870552063,
       "learning_rate": 1.6824800331324104e-06,
-      "loss": 1.8106,
+      "loss": 1.7319,
       "step": 30500
     },
     {
       "epoch": 0.612,
-      "grad_norm": 2.8769826889038086,
+      "grad_norm": 0.9811652898788452,
       "learning_rate": 1.6675022913315097e-06,
-      "loss": 1.8379,
+      "loss": 1.723,
       "step": 30600
     },
     {
       "epoch": 0.614,
-      "grad_norm": 2.66921067237854,
+      "grad_norm": 0.9791936278343201,
       "learning_rate": 1.6525580824169256e-06,
-      "loss": 1.8305,
+      "loss": 1.7305,
       "step": 30700
     },
     {
       "epoch": 0.616,
-      "grad_norm": 2.6944711208343506,
+      "grad_norm": 0.9365098476409912,
       "learning_rate": 1.6376480083392128e-06,
-      "loss": 1.8282,
+      "loss": 1.7326,
       "step": 30800
     },
     {
       "epoch": 0.618,
-      "grad_norm": 2.8102076053619385,
+      "grad_norm": 0.9214087724685669,
       "learning_rate": 1.6227726696739837e-06,
-      "loss": 1.8235,
+      "loss": 1.7279,
       "step": 30900
     },
     {
       "epoch": 0.62,
-      "grad_norm": 2.8156936168670654,
+      "grad_norm": 0.9731788635253906,
       "learning_rate": 1.6079326655977074e-06,
-      "loss": 1.8142,
+      "loss": 1.7216,
       "step": 31000
     },
     {
       "epoch": 0.62,
-      "eval_loss": 1.7180215120315552,
-      "eval_runtime": 24.4844,
-      "eval_samples_per_second": 204.212,
-      "eval_steps_per_second": 3.227,
+      "eval_loss": 1.7898601293563843,
+      "eval_runtime": 25.6025,
+      "eval_samples_per_second": 195.293,
+      "eval_steps_per_second": 3.086,
       "step": 31000
     },
     {
       "epoch": 0.622,
-      "grad_norm": 2.862630605697632,
+      "grad_norm": 0.956920862197876,
       "learning_rate": 1.5931285938635866e-06,
-      "loss": 1.8241,
+      "loss": 1.7295,
       "step": 31100
     },
     {
       "epoch": 0.624,
-      "grad_norm": 2.7605438232421875,
+      "grad_norm": 0.9402596950531006,
       "learning_rate": 1.5783610507774705e-06,
-      "loss": 1.8202,
+      "loss": 1.7241,
       "step": 31200
     },
     {
       "epoch": 0.626,
-      "grad_norm": 2.596388101577759,
+      "grad_norm": 0.9633392691612244,
       "learning_rate": 1.5636306311738425e-06,
-      "loss": 1.8329,
+      "loss": 1.7221,
       "step": 31300
     },
     {
       "epoch": 0.628,
-      "grad_norm": 2.9850687980651855,
+      "grad_norm": 0.9513702392578125,
       "learning_rate": 1.5489379283918566e-06,
-      "loss": 1.8301,
+      "loss": 1.7215,
       "step": 31400
     },
     {
       "epoch": 0.63,
-      "grad_norm": 2.886265754699707,
+      "grad_norm": 0.9428555369377136,
       "learning_rate": 1.534283534251437e-06,
-      "loss": 1.8439,
+      "loss": 1.7179,
       "step": 31500
     },
     {
       "epoch": 0.632,
-      "grad_norm": 2.7006332874298096,
+      "grad_norm": 0.9362800121307373,
       "learning_rate": 1.5196680390294437e-06,
-      "loss": 1.8371,
+      "loss": 1.7142,
       "step": 31600
     },
     {
       "epoch": 0.634,
-      "grad_norm": 2.658740520477295,
+      "grad_norm": 0.9473069906234741,
       "learning_rate": 1.5050920314358935e-06,
-      "loss": 1.8527,
+      "loss": 1.7232,
       "step": 31700
     },
     {
       "epoch": 0.636,
-      "grad_norm": 2.595576286315918,
+      "grad_norm": 0.9318504929542542,
       "learning_rate": 1.4905560985902472e-06,
-      "loss": 1.8319,
+      "loss": 1.7182,
       "step": 31800
     },
     {
       "epoch": 0.638,
-      "grad_norm": 2.9573771953582764,
+      "grad_norm": 0.939224123954773,
       "learning_rate": 1.4760608259977593e-06,
-      "loss": 1.8331,
+      "loss": 1.7209,
       "step": 31900
     },
     {
       "epoch": 0.64,
-      "grad_norm": 2.71058988571167,
+      "grad_norm": 0.9399820566177368,
       "learning_rate": 1.4616067975258964e-06,
-      "loss": 1.845,
+      "loss": 1.7143,
       "step": 32000
     },
     {
       "epoch": 0.64,
-      "eval_loss": 1.6738890409469604,
-      "eval_runtime": 24.4817,
-      "eval_samples_per_second": 204.234,
-      "eval_steps_per_second": 3.227,
+      "eval_loss": 1.7767833471298218,
+      "eval_runtime": 24.6525,
+      "eval_samples_per_second": 202.82,
+      "eval_steps_per_second": 3.205,
       "step": 32000
     },
     {
       "epoch": 0.642,
-      "grad_norm": 2.9249298572540283,
+      "grad_norm": 0.9213433265686035,
       "learning_rate": 1.447194595380818e-06,
-      "loss": 1.8374,
+      "loss": 1.7063,
       "step": 32100
     },
     {
       "epoch": 0.644,
-      "grad_norm": 2.7366347312927246,
+      "grad_norm": 0.9966316223144531,
       "learning_rate": 1.432824800083924e-06,
-      "loss": 1.8328,
+      "loss": 1.7227,
       "step": 32200
     },
     {
       "epoch": 0.646,
-      "grad_norm": 2.6436944007873535,
+      "grad_norm": 0.9814798831939697,
       "learning_rate": 1.4184979904484757e-06,
-      "loss": 1.819,
+      "loss": 1.7112,
       "step": 32300
     },
     {
       "epoch": 0.648,
-      "grad_norm": 3.2369704246520996,
+      "grad_norm": 0.9779679775238037,
       "learning_rate": 1.4042147435562754e-06,
-      "loss": 1.8154,
+      "loss": 1.72,
       "step": 32400
     },
     {
       "epoch": 0.65,
-      "grad_norm": 2.517576217651367,
+      "grad_norm": 0.9754687547683716,
       "learning_rate": 1.3899756347344235e-06,
-      "loss": 1.8284,
+      "loss": 1.6993,
       "step": 32500
     },
     {
       "epoch": 0.652,
-      "grad_norm": 3.261461019515991,
+      "grad_norm": 0.9322220683097839,
       "learning_rate": 1.3757812375321506e-06,
-      "loss": 1.8249,
+      "loss": 1.716,
       "step": 32600
     },
     {
       "epoch": 0.654,
-      "grad_norm": 3.1190226078033447,
+      "grad_norm": 0.9758723378181458,
       "learning_rate": 1.361632123697702e-06,
-      "loss": 1.8187,
+      "loss": 1.7164,
       "step": 32700
     },
     {
       "epoch": 0.656,
-      "grad_norm": 2.4606010913848877,
+      "grad_norm": 0.9605230093002319,
       "learning_rate": 1.3475288631553234e-06,
-      "loss": 1.8123,
+      "loss": 1.7122,
       "step": 32800
     },
     {
       "epoch": 0.658,
-      "grad_norm": 2.7457022666931152,
+      "grad_norm": 0.9641588926315308,
       "learning_rate": 1.3334720239822913e-06,
-      "loss": 1.8398,
+      "loss": 1.7133,
       "step": 32900
     },
     {
       "epoch": 0.66,
-      "grad_norm": 2.7495834827423096,
+      "grad_norm": 0.9363740682601929,
       "learning_rate": 1.3194621723860374e-06,
-      "loss": 1.84,
+      "loss": 1.7024,
       "step": 33000
     },
     {
       "epoch": 0.66,
-      "eval_loss": 1.7003716230392456,
-      "eval_runtime": 24.5107,
-      "eval_samples_per_second": 203.993,
-      "eval_steps_per_second": 3.223,
+      "eval_loss": 1.7694636583328247,
+      "eval_runtime": 24.6616,
+      "eval_samples_per_second": 202.745,
+      "eval_steps_per_second": 3.203,
       "step": 33000
     },
     {
       "epoch": 0.662,
-      "grad_norm": 2.70346999168396,
+      "grad_norm": 0.9194985628128052,
       "learning_rate": 1.3054998726813433e-06,
-      "loss": 1.8235,
+      "loss": 1.7087,
       "step": 33100
     },
     {
       "epoch": 0.664,
-      "grad_norm": 2.7724738121032715,
+      "grad_norm": 0.9364892244338989,
       "learning_rate": 1.2915856872676037e-06,
-      "loss": 1.8275,
+      "loss": 1.7146,
       "step": 33200
     },
     {
       "epoch": 0.666,
-      "grad_norm": 2.9709043502807617,
+      "grad_norm": 0.9384874701499939,
       "learning_rate": 1.2777201766061776e-06,
-      "loss": 1.8312,
+      "loss": 1.7026,
       "step": 33300
     },
     {
       "epoch": 0.668,
-      "grad_norm": 2.689776659011841,
+      "grad_norm": 0.9639517068862915,
       "learning_rate": 1.2639038991978141e-06,
-      "loss": 1.8454,
+      "loss": 1.7085,
       "step": 33400
     },
     {
       "epoch": 0.67,
-      "grad_norm": 2.7909674644470215,
+      "grad_norm": 0.9830998182296753,
       "learning_rate": 1.2501374115601522e-06,
-      "loss": 1.8301,
+      "loss": 1.7058,
       "step": 33500
     },
     {
       "epoch": 0.672,
-      "grad_norm": 2.7645773887634277,
+      "grad_norm": 0.9454180598258972,
       "learning_rate": 1.2364212682053068e-06,
-      "loss": 1.8387,
+      "loss": 1.7025,
       "step": 33600
     },
     {
       "epoch": 0.674,
-      "grad_norm": 2.7529468536376953,
+      "grad_norm": 0.9580062031745911,
       "learning_rate": 1.2227560216175317e-06,
-      "loss": 1.8214,
+      "loss": 1.7046,
       "step": 33700
     },
     {
       "epoch": 0.676,
-      "grad_norm": 2.870246410369873,
+      "grad_norm": 0.992702841758728,
       "learning_rate": 1.209142222230966e-06,
-      "loss": 1.8502,
+      "loss": 1.7037,
       "step": 33800
     },
     {
       "epoch": 0.678,
-      "grad_norm": 2.987135410308838,
+      "grad_norm": 0.9461750388145447,
       "learning_rate": 1.195580418407466e-06,
-      "loss": 1.8398,
+      "loss": 1.6995,
       "step": 33900
     },
     {
       "epoch": 0.68,
-      "grad_norm": 2.5931384563446045,
+      "grad_norm": 0.9390709400177002,
       "learning_rate": 1.1820711564145109e-06,
-      "loss": 1.8008,
+      "loss": 1.6926,
       "step": 34000
     },
     {
       "epoch": 0.68,
-      "eval_loss": 1.7073742151260376,
-      "eval_runtime": 24.5455,
-      "eval_samples_per_second": 203.703,
-      "eval_steps_per_second": 3.219,
+      "eval_loss": 1.8035398721694946,
+      "eval_runtime": 24.6976,
+      "eval_samples_per_second": 202.449,
+      "eval_steps_per_second": 3.199,
       "step": 34000
     },
     {
       "epoch": 0.682,
-      "grad_norm": 2.6862409114837646,
+      "grad_norm": 0.9636568427085876,
       "learning_rate": 1.1686149804032038e-06,
-      "loss": 1.8288,
+      "loss": 1.6947,
       "step": 34100
     },
     {
       "epoch": 0.684,
-      "grad_norm": 2.8123841285705566,
+      "grad_norm": 0.9337140917778015,
       "learning_rate": 1.1552124323863537e-06,
-      "loss": 1.8282,
+      "loss": 1.6931,
       "step": 34200
     },
     {
       "epoch": 0.686,
-      "grad_norm": 2.80928635597229,
+      "grad_norm": 0.9221658110618591,
       "learning_rate": 1.1418640522166393e-06,
-      "loss": 1.8372,
+      "loss": 1.7008,
       "step": 34300
     },
     {
       "epoch": 0.688,
-      "grad_norm": 2.4395246505737305,
+      "grad_norm": 0.9931758046150208,
       "learning_rate": 1.1285703775648682e-06,
-      "loss": 1.8196,
+      "loss": 1.6919,
       "step": 34400
     },
     {
       "epoch": 0.69,
-      "grad_norm": 2.514472723007202,
+      "grad_norm": 0.9167136549949646,
       "learning_rate": 1.1153319438983162e-06,
-      "loss": 1.8219,
+      "loss": 1.6957,
       "step": 34500
     },
     {
       "epoch": 0.692,
-      "grad_norm": 2.937126874923706,
+      "grad_norm": 0.9322009086608887,
       "learning_rate": 1.1021492844591605e-06,
-      "loss": 1.8162,
+      "loss": 1.6946,
       "step": 34600
     },
     {
       "epoch": 0.694,
-      "grad_norm": 2.6019275188446045,
+      "grad_norm": 0.9542787075042725,
       "learning_rate": 1.0890229302430028e-06,
-      "loss": 1.8315,
+      "loss": 1.7092,
       "step": 34700
     },
     {
       "epoch": 0.696,
-      "grad_norm": 2.926776647567749,
+      "grad_norm": 0.9714674949645996,
       "learning_rate": 1.0759534099774763e-06,
-      "loss": 1.8323,
+      "loss": 1.6895,
       "step": 34800
     },
     {
       "epoch": 0.698,
-      "grad_norm": 2.9355075359344482,
+      "grad_norm": 0.9719817042350769,
       "learning_rate": 1.0629412501009504e-06,
-      "loss": 1.835,
+      "loss": 1.6933,
       "step": 34900
     },
     {
       "epoch": 0.7,
-      "grad_norm": 2.810500144958496,
+      "grad_norm": 0.921370267868042,
       "learning_rate": 1.0499869747413295e-06,
-      "loss": 1.8333,
+      "loss": 1.6885,
       "step": 35000
     },
     {
       "epoch": 0.7,
-      "eval_loss": 1.6875180006027222,
-      "eval_runtime": 24.5448,
-      "eval_samples_per_second": 203.71,
-      "eval_steps_per_second": 3.219,
+      "eval_loss": 1.769335150718689,
+      "eval_runtime": 24.6974,
+      "eval_samples_per_second": 202.45,
+      "eval_steps_per_second": 3.199,
       "step": 35000
     },
     {
       "epoch": 0.702,
-      "grad_norm": 2.754328727722168,
+      "grad_norm": 0.887254536151886,
       "learning_rate": 1.0370911056949353e-06,
-      "loss": 1.8135,
+      "loss": 1.6882,
       "step": 35100
     },
     {
       "epoch": 0.704,
-      "grad_norm": 2.738758087158203,
+      "grad_norm": 0.9177955985069275,
       "learning_rate": 1.0242541624054925e-06,
-      "loss": 1.8299,
+      "loss": 1.693,
       "step": 35200
     },
     {
       "epoch": 0.706,
-      "grad_norm": 2.985809564590454,
+      "grad_norm": 0.9564446210861206,
       "learning_rate": 1.0114766619432052e-06,
-      "loss": 1.8395,
+      "loss": 1.6849,
       "step": 35300
     },
     {
       "epoch": 0.708,
-      "grad_norm": 2.5629913806915283,
+      "grad_norm": 0.9259414672851562,
       "learning_rate": 9.987591189839277e-07,
-      "loss": 1.8024,
+      "loss": 1.6953,
       "step": 35400
     },
     {
       "epoch": 0.71,
-      "grad_norm": 2.4209351539611816,
+      "grad_norm": 1.041141390800476,
       "learning_rate": 9.861020457884383e-07,
-      "loss": 1.805,
+      "loss": 1.6952,
       "step": 35500
     },
     {
       "epoch": 0.712,
-      "grad_norm": 3.0288236141204834,
+      "grad_norm": 0.941935122013092,
       "learning_rate": 9.735059521817983e-07,
-      "loss": 1.8175,
+      "loss": 1.6901,
       "step": 35600
     },
     {
       "epoch": 0.714,
-      "grad_norm": 2.89047908782959,
+      "grad_norm": 0.9669387340545654,
       "learning_rate": 9.60971345532821e-07,
-      "loss": 1.8144,
+      "loss": 1.6816,
       "step": 35700
     },
     {
       "epoch": 0.716,
-      "grad_norm": 2.7108571529388428,
+      "grad_norm": 0.9418676495552063,
       "learning_rate": 9.48498730733636e-07,
-      "loss": 1.8238,
+      "loss": 1.6881,
       "step": 35800
     },
     {
       "epoch": 0.718,
-      "grad_norm": 2.547468423843384,
+      "grad_norm": 0.9330548644065857,
       "learning_rate": 9.360886101793481e-07,
-      "loss": 1.82,
+      "loss": 1.6835,
       "step": 35900
     },
     {
       "epoch": 0.72,
-      "grad_norm": 2.698509931564331,
+      "grad_norm": 0.9843077659606934,
       "learning_rate": 9.237414837478026e-07,
-      "loss": 1.8229,
+      "loss": 1.6829,
       "step": 36000
     },
     {
       "epoch": 0.72,
-      "eval_loss": 1.7106120586395264,
-      "eval_runtime": 24.5868,
-      "eval_samples_per_second": 203.361,
-      "eval_steps_per_second": 3.213,
+      "eval_loss": 1.8024219274520874,
+      "eval_runtime": 24.6857,
+      "eval_samples_per_second": 202.546,
+      "eval_steps_per_second": 3.2,
       "step": 36000
     },
     {
       "epoch": 0.722,
-      "grad_norm": 2.7523186206817627,
+      "grad_norm": 0.9610580801963806,
       "learning_rate": 9.114578487794545e-07,
-      "loss": 1.8365,
+      "loss": 1.686,
       "step": 36100
     },
     {
       "epoch": 0.724,
-      "grad_norm": 2.7361319065093994,
+      "grad_norm": 0.9139875769615173,
       "learning_rate": 8.992382000573258e-07,
-      "loss": 1.816,
+      "loss": 1.6858,
       "step": 36200
     },
     {
       "epoch": 0.726,
-      "grad_norm": 2.705658435821533,
+      "grad_norm": 0.9467193484306335,
       "learning_rate": 8.870830297870878e-07,
-      "loss": 1.8239,
+      "loss": 1.6792,
       "step": 36300
     },
     {
       "epoch": 0.728,
-      "grad_norm": 2.538872480392456,
+      "grad_norm": 0.9271172881126404,
       "learning_rate": 8.749928275772256e-07,
-      "loss": 1.8157,
+      "loss": 1.682,
       "step": 36400
     },
     {
       "epoch": 0.73,
-      "grad_norm": 2.850130081176758,
+      "grad_norm": 0.9212483167648315,
       "learning_rate": 8.629680804193208e-07,
-      "loss": 1.8243,
+      "loss": 1.6871,
       "step": 36500
     },
     {
       "epoch": 0.732,
-      "grad_norm": 2.8845088481903076,
+      "grad_norm": 0.9672465920448303,
       "learning_rate": 8.510092726684368e-07,
-      "loss": 1.8352,
+      "loss": 1.6959,
       "step": 36600
     },
     {
       "epoch": 0.734,
-      "grad_norm": 2.4896080493927,
+      "grad_norm": 0.9895572662353516,
       "learning_rate": 8.391168860236057e-07,
-      "loss": 1.8174,
+      "loss": 1.6851,
       "step": 36700
     },
     {
       "epoch": 0.736,
-      "grad_norm": 2.7589054107666016,
+      "grad_norm": 0.9622398018836975,
       "learning_rate": 8.272913995084262e-07,
-      "loss": 1.8154,
+      "loss": 1.685,
       "step": 36800
     },
     {
       "epoch": 0.738,
-      "grad_norm": 2.98602032661438,
+      "grad_norm": 0.9831019043922424,
       "learning_rate": 8.155332894517734e-07,
-      "loss": 1.8333,
+      "loss": 1.6854,
       "step": 36900
     },
     {
       "epoch": 0.74,
-      "grad_norm": 2.650705099105835,
+      "grad_norm": 0.9320428967475891,
       "learning_rate": 8.038430294686028e-07,
-      "loss": 1.8321,
+      "loss": 1.6775,
       "step": 37000
     },
     {
       "epoch": 0.74,
-      "eval_loss": 1.6874617338180542,
-      "eval_runtime": 24.555,
-      "eval_samples_per_second": 203.625,
-      "eval_steps_per_second": 3.217,
+      "eval_loss": 1.773045539855957,
+      "eval_runtime": 24.7195,
+      "eval_samples_per_second": 202.269,
+      "eval_steps_per_second": 3.196,
       "step": 37000
     },
     {
       "epoch": 0.742,
-      "grad_norm": 2.9605143070220947,
+      "grad_norm": 0.9403007626533508,
       "learning_rate": 7.922210904408844e-07,
-      "loss": 1.8179,
+      "loss": 1.6863,
       "step": 37100
     },
     {
       "epoch": 0.744,
-      "grad_norm": 2.5343451499938965,
+      "grad_norm": 0.9451361298561096,
       "learning_rate": 7.806679404986264e-07,
-      "loss": 1.8119,
+      "loss": 1.6787,
       "step": 37200
     },
     {
       "epoch": 0.746,
-      "grad_norm": 2.6719601154327393,
+      "grad_norm": 0.9065183997154236,
       "learning_rate": 7.691840450010238e-07,
-      "loss": 1.8277,
+      "loss": 1.674,
       "step": 37300
     },
     {
       "epoch": 0.748,
-      "grad_norm": 2.680201530456543,
+      "grad_norm": 0.9662121534347534,
       "learning_rate": 7.577698665177136e-07,
-      "loss": 1.8013,
+      "loss": 1.6806,
       "step": 37400
     },
     {
       "epoch": 0.75,
-      "grad_norm": 2.7274270057678223,
+      "grad_norm": 0.9532101154327393,
       "learning_rate": 7.464258648101402e-07,
-      "loss": 1.8035,
+      "loss": 1.6854,
       "step": 37500
     },
     {
       "epoch": 0.752,
-      "grad_norm": 2.7913053035736084,
+      "grad_norm": 0.930415689945221,
       "learning_rate": 7.351524968130377e-07,
-      "loss": 1.8227,
+      "loss": 1.6865,
       "step": 37600
     },
     {
       "epoch": 0.754,
-      "grad_norm": 2.7317264080047607,
+      "grad_norm": 0.9601901173591614,
       "learning_rate": 7.239502166160267e-07,
-      "loss": 1.8135,
+      "loss": 1.6752,
       "step": 37700
     },
     {
       "epoch": 0.756,
-      "grad_norm": 2.7197861671447754,
+      "grad_norm": 0.9325251579284668,
       "learning_rate": 7.128194754453191e-07,
-      "loss": 1.8272,
+      "loss": 1.6714,
       "step": 37800
     },
     {
       "epoch": 0.758,
-      "grad_norm": 2.622196912765503,
+      "grad_norm": 0.9412240386009216,
       "learning_rate": 7.017607216455466e-07,
-      "loss": 1.8283,
+      "loss": 1.6788,
       "step": 37900
     },
     {
       "epoch": 0.76,
-      "grad_norm": 2.676344633102417,
+      "grad_norm": 0.9871384501457214,
       "learning_rate": 6.907744006616993e-07,
-      "loss": 1.8327,
+      "loss": 1.6819,
       "step": 38000
     },
     {
       "epoch": 0.76,
-      "eval_loss": 1.6959562301635742,
-      "eval_runtime": 24.5554,
-      "eval_samples_per_second": 203.621,
-      "eval_steps_per_second": 3.217,
+      "eval_loss": 1.7901370525360107,
+      "eval_runtime": 24.7095,
+      "eval_samples_per_second": 202.351,
+      "eval_steps_per_second": 3.197,
       "step": 38000
     },
     {
       "epoch": 0.762,
-      "grad_norm": 2.644090414047241,
+      "grad_norm": 0.9582891464233398,
       "learning_rate": 6.798609550211844e-07,
-      "loss": 1.8168,
+      "loss": 1.6761,
       "step": 38100
     },
     {
       "epoch": 0.764,
-      "grad_norm": 2.7100324630737305,
+      "grad_norm": 0.9154219627380371,
       "learning_rate": 6.690208243160015e-07,
-      "loss": 1.8316,
+      "loss": 1.6818,
       "step": 38200
     },
     {
       "epoch": 0.766,
-      "grad_norm": 2.867020845413208,
+      "grad_norm": 0.9294545650482178,
       "learning_rate": 6.58254445185035e-07,
-      "loss": 1.818,
+      "loss": 1.6758,
       "step": 38300
     },
     {
       "epoch": 0.768,
-      "grad_norm": 2.7859420776367188,
+      "grad_norm": 0.9826803803443909,
       "learning_rate": 6.475622512964653e-07,
-      "loss": 1.8221,
+      "loss": 1.6758,
       "step": 38400
     },
     {
       "epoch": 0.77,
-      "grad_norm": 2.8102128505706787,
+      "grad_norm": 0.9328571557998657,
       "learning_rate": 6.369446733303045e-07,
-      "loss": 1.8284,
+      "loss": 1.6797,
       "step": 38500
     },
     {
       "epoch": 0.772,
-      "grad_norm": 2.837434768676758,
+      "grad_norm": 0.9535841345787048,
       "learning_rate": 6.26402138961045e-07,
-      "loss": 1.8373,
+      "loss": 1.681,
       "step": 38600
     },
     {
       "epoch": 0.774,
-      "grad_norm": 3.005826473236084,
+      "grad_norm": 0.9459956288337708,
       "learning_rate": 6.159350728404318e-07,
-      "loss": 1.8335,
+      "loss": 1.6712,
       "step": 38700
     },
     {
       "epoch": 0.776,
-      "grad_norm": 2.7674293518066406,
+      "grad_norm": 0.931379497051239,
       "learning_rate": 6.05543896580364e-07,
-      "loss": 1.8177,
+      "loss": 1.6747,
       "step": 38800
     },
     {
       "epoch": 0.778,
-      "grad_norm": 2.7954766750335693,
+      "grad_norm": 0.9323126077651978,
       "learning_rate": 5.95229028735905e-07,
-      "loss": 1.8304,
+      "loss": 1.68,
       "step": 38900
     },
     {
       "epoch": 0.78,
-      "grad_norm": 2.581066370010376,
+      "grad_norm": 0.9810007810592651,
       "learning_rate": 5.849908847884289e-07,
-      "loss": 1.8336,
+      "loss": 1.6846,
       "step": 39000
     },
     {
       "epoch": 0.78,
-      "eval_loss": 1.6949576139450073,
-      "eval_runtime": 24.562,
-      "eval_samples_per_second": 203.566,
-      "eval_steps_per_second": 3.216,
+      "eval_loss": 1.8078017234802246,
+      "eval_runtime": 24.7368,
+      "eval_samples_per_second": 202.128,
+      "eval_steps_per_second": 3.194,
       "step": 39000
     },
     {
       "epoch": 0.782,
-      "grad_norm": 2.7293710708618164,
+      "grad_norm": 0.9599342942237854,
       "learning_rate": 5.748298771288801e-07,
-      "loss": 1.822,
+      "loss": 1.6816,
       "step": 39100
     },
     {
       "epoch": 0.784,
-      "grad_norm": 2.5825986862182617,
+      "grad_norm": 0.9463416337966919,
       "learning_rate": 5.647464150411655e-07,
-      "loss": 1.8298,
+      "loss": 1.6707,
       "step": 39200
     },
     {
       "epoch": 0.786,
-      "grad_norm": 2.420144557952881,
+      "grad_norm": 0.9661208391189575,
       "learning_rate": 5.547409046856683e-07,
-      "loss": 1.8071,
+      "loss": 1.6706,
       "step": 39300
     },
     {
       "epoch": 0.788,
-      "grad_norm": 2.877675771713257,
+      "grad_norm": 0.8846033811569214,
       "learning_rate": 5.448137490828862e-07,
-      "loss": 1.8203,
+      "loss": 1.6755,
       "step": 39400
     },
     {
       "epoch": 0.79,
-      "grad_norm": 2.6774392127990723,
+      "grad_norm": 0.9732816219329834,
       "learning_rate": 5.349653480971997e-07,
-      "loss": 1.8286,
+      "loss": 1.6756,
       "step": 39500
     },
     {
       "epoch": 0.792,
-      "grad_norm": 2.545191526412964,
+      "grad_norm": 0.955687403678894,
       "learning_rate": 5.25196098420764e-07,
-      "loss": 1.817,
+      "loss": 1.683,
       "step": 39600
     },
     {
       "epoch": 0.794,
-      "grad_norm": 2.946167469024658,
+      "grad_norm": 1.00485098361969,
       "learning_rate": 5.155063935575322e-07,
-      "loss": 1.8259,
+      "loss": 1.6808,
       "step": 39700
     },
     {
       "epoch": 0.796,
-      "grad_norm": 2.8756895065307617,
+      "grad_norm": 0.9409605860710144,
       "learning_rate": 5.05896623807402e-07,
-      "loss": 1.827,
+      "loss": 1.6775,
       "step": 39800
     },
     {
       "epoch": 0.798,
-      "grad_norm": 2.8700594902038574,
+      "grad_norm": 0.959143340587616,
       "learning_rate": 4.963671762504993e-07,
-      "loss": 1.8209,
+      "loss": 1.6787,
       "step": 39900
     },
     {
       "epoch": 0.8,
-      "grad_norm": 2.6594278812408447,
+      "grad_norm": 0.9962793588638306,
       "learning_rate": 4.869184347315811e-07,
-      "loss": 1.8385,
+      "loss": 1.6805,
       "step": 40000
     },
     {
       "epoch": 0.8,
-      "eval_loss": 1.6786351203918457,
-      "eval_runtime": 24.5618,
-      "eval_samples_per_second": 203.568,
-      "eval_steps_per_second": 3.216,
+      "eval_loss": 1.7852710485458374,
+      "eval_runtime": 24.7601,
+      "eval_samples_per_second": 201.938,
+      "eval_steps_per_second": 3.191,
       "step": 40000
     },
     {
       "epoch": 0.802,
-      "grad_norm": 2.7170181274414062,
+      "grad_norm": 0.9732511639595032,
       "learning_rate": 4.775507798445788e-07,
-      "loss": 1.8099,
+      "loss": 1.668,
       "step": 40100
     },
     {
       "epoch": 0.804,
-      "grad_norm": 2.908679246902466,
+      "grad_norm": 0.9450587630271912,
       "learning_rate": 4.6826458891726517e-07,
-      "loss": 1.8184,
+      "loss": 1.6811,
       "step": 40200
     },
     {
       "epoch": 0.806,
-      "grad_norm": 2.8288190364837646,
+      "grad_norm": 0.9879899024963379,
       "learning_rate": 4.5906023599605663e-07,
-      "loss": 1.8182,
+      "loss": 1.6672,
       "step": 40300
     },
     {
       "epoch": 0.808,
-      "grad_norm": 2.8533594608306885,
+      "grad_norm": 0.9791668653488159,
       "learning_rate": 4.499380918309465e-07,
-      "loss": 1.8056,
+      "loss": 1.6715,
       "step": 40400
     },
     {
       "epoch": 0.81,
-      "grad_norm": 2.694812774658203,
+      "grad_norm": 0.9480279684066772,
       "learning_rate": 4.4089852386057174e-07,
-      "loss": 1.8186,
+      "loss": 1.6701,
       "step": 40500
     },
     {
       "epoch": 0.812,
-      "grad_norm": 2.7870705127716064,
+      "grad_norm": 0.9733260869979858,
       "learning_rate": 4.3194189619741144e-07,
-      "loss": 1.791,
+      "loss": 1.6692,
       "step": 40600
     },
     {
       "epoch": 0.814,
-      "grad_norm": 2.5385093688964844,
+      "grad_norm": 1.0035480260849,
       "learning_rate": 4.2306856961312314e-07,
-      "loss": 1.8046,
+      "loss": 1.6787,
       "step": 40700
     },
     {
       "epoch": 0.816,
-      "grad_norm": 2.468137741088867,
+      "grad_norm": 0.9302639961242676,
       "learning_rate": 4.142789015240073e-07,
-      "loss": 1.7792,
+      "loss": 1.6736,
       "step": 40800
     },
     {
       "epoch": 0.818,
-      "grad_norm": 2.8535079956054688,
+      "grad_norm": 0.9349082112312317,
       "learning_rate": 4.0557324597661233e-07,
-      "loss": 1.8112,
+      "loss": 1.6703,
       "step": 40900
     },
     {
       "epoch": 0.82,
-      "grad_norm": 2.3535666465759277,
+      "grad_norm": 0.954745888710022,
       "learning_rate": 3.969519536334751e-07,
-      "loss": 1.8047,
+      "loss": 1.6688,
       "step": 41000
     },
     {
       "epoch": 0.82,
-      "eval_loss": 1.686981201171875,
-      "eval_runtime": 25.5497,
-      "eval_samples_per_second": 195.697,
-      "eval_steps_per_second": 3.092,
+      "eval_loss": 1.794049859046936,
+      "eval_runtime": 24.8218,
+      "eval_samples_per_second": 201.436,
+      "eval_steps_per_second": 3.183,
       "step": 41000
     },
     {
       "epoch": 0.822,
-      "grad_norm": 2.6770412921905518,
+      "grad_norm": 1.0045809745788574,
       "learning_rate": 3.8841537175899364e-07,
-      "loss": 1.7946,
+      "loss": 1.6669,
       "step": 41100
     },
     {
       "epoch": 0.824,
-      "grad_norm": 2.5880556106567383,
+      "grad_norm": 0.9524197578430176,
       "learning_rate": 3.7996384420544206e-07,
-      "loss": 1.7951,
+      "loss": 1.6581,
       "step": 41200
     },
     {
       "epoch": 0.826,
-      "grad_norm": 2.679513931274414,
+      "grad_norm": 0.9386103749275208,
       "learning_rate": 3.715977113991165e-07,
-      "loss": 1.7906,
+      "loss": 1.6756,
       "step": 41300
     },
     {
       "epoch": 0.828,
-      "grad_norm": 2.9302854537963867,
+      "grad_norm": 0.9285281300544739,
       "learning_rate": 3.633173103266266e-07,
-      "loss": 1.8058,
+      "loss": 1.666,
       "step": 41400
     },
     {
       "epoch": 0.83,
-      "grad_norm": 2.778660297393799,
+      "grad_norm": 0.9478991031646729,
       "learning_rate": 3.551229745213211e-07,
-      "loss": 1.8033,
+      "loss": 1.6651,
       "step": 41500
     },
     {
       "epoch": 0.832,
-      "grad_norm": 2.921802043914795,
+      "grad_norm": 0.9116945862770081,
       "learning_rate": 3.47015034049851e-07,
-      "loss": 1.803,
+      "loss": 1.686,
       "step": 41600
     },
     {
       "epoch": 0.834,
-      "grad_norm": 2.771888256072998,
+      "grad_norm": 0.9529798030853271,
       "learning_rate": 3.3899381549887563e-07,
-      "loss": 1.806,
+      "loss": 1.6834,
       "step": 41700
     },
     {
       "epoch": 0.836,
-      "grad_norm": 2.689772367477417,
+      "grad_norm": 0.9535943865776062,
       "learning_rate": 3.310596419619097e-07,
-      "loss": 1.7827,
+      "loss": 1.6702,
       "step": 41800
     },
     {
       "epoch": 0.838,
-      "grad_norm": 2.4702703952789307,
+      "grad_norm": 0.9451871514320374,
       "learning_rate": 3.23212833026306e-07,
-      "loss": 1.7785,
+      "loss": 1.6731,
       "step": 41900
     },
     {
       "epoch": 0.84,
-      "grad_norm": 2.8901641368865967,
+      "grad_norm": 0.9637697339057922,
       "learning_rate": 3.1545370476038394e-07,
-      "loss": 1.7997,
+      "loss": 1.6629,
       "step": 42000
     },
     {
       "epoch": 0.84,
-      "eval_loss": 1.6959081888198853,
-      "eval_runtime": 24.5827,
-      "eval_samples_per_second": 203.395,
-      "eval_steps_per_second": 3.214,
+      "eval_loss": 1.792710781097412,
+      "eval_runtime": 24.7667,
+      "eval_samples_per_second": 201.884,
+      "eval_steps_per_second": 3.19,
       "step": 42000
     },
     {
       "epoch": 0.842,
-      "grad_norm": 2.830742359161377,
+      "grad_norm": 0.9361735582351685,
       "learning_rate": 3.077825697006989e-07,
-      "loss": 1.7988,
+      "loss": 1.6698,
       "step": 42100
     },
     {
       "epoch": 0.844,
-      "grad_norm": 2.8089168071746826,
+      "grad_norm": 0.9601408839225769,
       "learning_rate": 3.0019973683945216e-07,
-      "loss": 1.7862,
+      "loss": 1.6786,
       "step": 42200
     },
     {
       "epoch": 0.846,
-      "grad_norm": 2.7035460472106934,
+      "grad_norm": 0.9747889041900635,
       "learning_rate": 2.927055116120464e-07,
-      "loss": 1.7711,
+      "loss": 1.6653,
       "step": 42300
     },
     {
       "epoch": 0.848,
-      "grad_norm": 2.6916584968566895,
+      "grad_norm": 0.9499157667160034,
       "learning_rate": 2.8530019588478093e-07,
-      "loss": 1.8009,
+      "loss": 1.6872,
       "step": 42400
     },
     {
       "epoch": 0.85,
-      "grad_norm": 2.679772138595581,
+      "grad_norm": 0.9127237200737,
       "learning_rate": 2.779840879426923e-07,
-      "loss": 1.8032,
+      "loss": 1.6667,
       "step": 42500
     },
     {
       "epoch": 0.852,
-      "grad_norm": 2.703826665878296,
+      "grad_norm": 0.9435052275657654,
       "learning_rate": 2.707574824775433e-07,
-      "loss": 1.8069,
+      "loss": 1.666,
       "step": 42600
     },
     {
       "epoch": 0.854,
-      "grad_norm": 2.61206316947937,
+      "grad_norm": 0.9535505175590515,
       "learning_rate": 2.6362067057594754e-07,
-      "loss": 1.7922,
+      "loss": 1.6744,
       "step": 42700
     },
     {
       "epoch": 0.856,
-      "grad_norm": 2.7829911708831787,
+      "grad_norm": 0.930338978767395,
       "learning_rate": 2.565739397076475e-07,
-      "loss": 1.7881,
+      "loss": 1.665,
       "step": 42800
     },
     {
       "epoch": 0.858,
-      "grad_norm": 2.5335564613342285,
+      "grad_norm": 0.9728429317474365,
       "learning_rate": 2.4961757371393614e-07,
-      "loss": 1.8012,
+      "loss": 1.6635,
       "step": 42900
     },
     {
       "epoch": 0.86,
-      "grad_norm": 2.6119823455810547,
+      "grad_norm": 0.9958027601242065,
       "learning_rate": 2.427518527962194e-07,
-      "loss": 1.8008,
+      "loss": 1.6721,
       "step": 43000
     },
     {
       "epoch": 0.86,
-      "eval_loss": 1.6836472749710083,
-      "eval_runtime": 24.6165,
-      "eval_samples_per_second": 203.116,
-      "eval_steps_per_second": 3.209,
+      "eval_loss": 1.7779754400253296,
+      "eval_runtime": 24.7675,
+      "eval_samples_per_second": 201.878,
+      "eval_steps_per_second": 3.19,
       "step": 43000
     },
     {
       "epoch": 0.862,
-      "grad_norm": 2.7010042667388916,
+      "grad_norm": 0.9844805002212524,
       "learning_rate": 2.3597705350473677e-07,
-      "loss": 1.787,
+      "loss": 1.6651,
       "step": 43100
     },
     {
       "epoch": 0.864,
-      "grad_norm": 2.5886828899383545,
+      "grad_norm": 0.9666577577590942,
       "learning_rate": 2.2929344872741482e-07,
-      "loss": 1.8017,
+      "loss": 1.6636,
       "step": 43200
     },
     {
       "epoch": 0.866,
-      "grad_norm": 2.6632063388824463,
+      "grad_norm": 0.9284493923187256,
       "learning_rate": 2.227013076788792e-07,
-      "loss": 1.804,
+      "loss": 1.6684,
       "step": 43300
     },
     {
       "epoch": 0.868,
-      "grad_norm": 2.5856895446777344,
+      "grad_norm": 0.9273450970649719,
       "learning_rate": 2.1620089588961084e-07,
-      "loss": 1.7809,
+      "loss": 1.6667,
       "step": 43400
     },
     {
       "epoch": 0.87,
-      "grad_norm": 2.8796870708465576,
+      "grad_norm": 0.9567154049873352,
       "learning_rate": 2.0979247519524875e-07,
-      "loss": 1.7951,
+      "loss": 1.6762,
       "step": 43500
     },
     {
       "epoch": 0.872,
-      "grad_norm": 2.6709506511688232,
+      "grad_norm": 0.9734944701194763,
       "learning_rate": 2.0347630372604316e-07,
-      "loss": 1.7951,
+      "loss": 1.6701,
       "step": 43600
     },
     {
       "epoch": 0.874,
-      "grad_norm": 2.7470643520355225,
+      "grad_norm": 0.9483113884925842,
       "learning_rate": 1.9725263589646132e-07,
-      "loss": 1.7909,
+      "loss": 1.6593,
       "step": 43700
     },
     {
       "epoch": 0.876,
-      "grad_norm": 2.5582199096679688,
+      "grad_norm": 0.9375894069671631,
       "learning_rate": 1.911217223949341e-07,
-      "loss": 1.8007,
+      "loss": 1.6765,
       "step": 43800
     },
     {
       "epoch": 0.878,
-      "grad_norm": 2.737184524536133,
+      "grad_norm": 0.9671183228492737,
       "learning_rate": 1.850838101737648e-07,
-      "loss": 1.782,
+      "loss": 1.6711,
       "step": 43900
     },
     {
       "epoch": 0.88,
-      "grad_norm": 2.5171310901641846,
+      "grad_norm": 0.9503777623176575,
       "learning_rate": 1.7913914243917696e-07,
-      "loss": 1.7803,
+      "loss": 1.6752,
       "step": 44000
     },
     {
       "epoch": 0.88,
-      "eval_loss": 1.6784179210662842,
-      "eval_runtime": 24.6165,
-      "eval_samples_per_second": 203.115,
-      "eval_steps_per_second": 3.209,
+      "eval_loss": 1.7995929718017578,
+      "eval_runtime": 24.7688,
+      "eval_samples_per_second": 201.867,
+      "eval_steps_per_second": 3.189,
       "step": 44000
     },
     {
       "epoch": 0.882,
-      "grad_norm": 2.7506484985351562,
+      "grad_norm": 0.9515755772590637,
       "learning_rate": 1.7328795864151953e-07,
-      "loss": 1.775,
+      "loss": 1.6649,
       "step": 44100
     },
     {
       "epoch": 0.884,
-      "grad_norm": 2.688084363937378,
+      "grad_norm": 0.9477518200874329,
       "learning_rate": 1.6753049446562363e-07,
-      "loss": 1.7881,
+      "loss": 1.6761,
       "step": 44200
     },
     {
       "epoch": 0.886,
-      "grad_norm": 2.5896782875061035,
+      "grad_norm": 0.9282898306846619,
       "learning_rate": 1.6186698182130645e-07,
-      "loss": 1.7941,
+      "loss": 1.6684,
       "step": 44300
     },
     {
       "epoch": 0.888,
-      "grad_norm": 2.6385583877563477,
+      "grad_norm": 0.972673773765564,
       "learning_rate": 1.5629764883403066e-07,
-      "loss": 1.7966,
+      "loss": 1.6697,
       "step": 44400
     },
     {
       "epoch": 0.89,
-      "grad_norm": 2.5691211223602295,
+      "grad_norm": 0.9413368701934814,
       "learning_rate": 1.5082271983571816e-07,
-      "loss": 1.7908,
+      "loss": 1.6711,
       "step": 44500
     },
     {
       "epoch": 0.892,
-      "grad_norm": 2.924225091934204,
+      "grad_norm": 0.9665236473083496,
       "learning_rate": 1.454424153557099e-07,
-      "loss": 1.783,
+      "loss": 1.671,
       "step": 44600
     },
     {
       "epoch": 0.894,
-      "grad_norm": 2.4737792015075684,
+      "grad_norm": 0.9585925340652466,
       "learning_rate": 1.4015695211188618e-07,
-      "loss": 1.7977,
+      "loss": 1.6715,
       "step": 44700
     },
     {
       "epoch": 0.896,
-      "grad_norm": 2.56907320022583,
+      "grad_norm": 0.9342137575149536,
       "learning_rate": 1.3496654300193534e-07,
-      "loss": 1.7903,
+      "loss": 1.6691,
       "step": 44800
     },
     {
       "epoch": 0.898,
-      "grad_norm": 2.9577035903930664,
+      "grad_norm": 0.9924020767211914,
       "learning_rate": 1.298713970947796e-07,
-      "loss": 1.8048,
+      "loss": 1.6735,
       "step": 44900
     },
     {
       "epoch": 0.9,
-      "grad_norm": 2.7337706089019775,
+      "grad_norm": 0.9722588658332825,
       "learning_rate": 1.2487171962215382e-07,
-      "loss": 1.7927,
+      "loss": 1.6684,
       "step": 45000
     },
     {
       "epoch": 0.9,
-      "eval_loss": 1.6848783493041992,
-      "eval_runtime": 24.656,
-      "eval_samples_per_second": 202.79,
-      "eval_steps_per_second": 3.204,
+      "eval_loss": 1.793320894241333,
+      "eval_runtime": 24.8095,
+      "eval_samples_per_second": 201.535,
+      "eval_steps_per_second": 3.184,
       "step": 45000
     },
     {
       "epoch": 0.902,
-      "grad_norm": 2.6311519145965576,
+      "grad_norm": 0.9450165033340454,
       "learning_rate": 1.1996771197033685e-07,
-      "loss": 1.7955,
+      "loss": 1.6621,
       "step": 45100
     },
     {
       "epoch": 0.904,
-      "grad_norm": 2.556652307510376,
+      "grad_norm": 0.9521385431289673,
       "learning_rate": 1.151595716720419e-07,
-      "loss": 1.788,
+      "loss": 1.6771,
       "step": 45200
     },
     {
       "epoch": 0.906,
-      "grad_norm": 2.887894868850708,
+      "grad_norm": 0.9356434941291809,
       "learning_rate": 1.1044749239845942e-07,
-      "loss": 1.8029,
+      "loss": 1.6645,
       "step": 45300
     },
     {
       "epoch": 0.908,
-      "grad_norm": 3.0363547801971436,
+      "grad_norm": 0.9615120887756348,
       "learning_rate": 1.0583166395145533e-07,
-      "loss": 1.7893,
+      "loss": 1.6596,
       "step": 45400
     },
     {
       "epoch": 0.91,
-      "grad_norm": 2.726654052734375,
+      "grad_norm": 0.9194362759590149,
       "learning_rate": 1.0131227225592649e-07,
-      "loss": 1.7937,
+      "loss": 1.6653,
       "step": 45500
     },
     {
       "epoch": 0.912,
-      "grad_norm": 2.5315470695495605,
+      "grad_norm": 0.9323084354400635,
       "learning_rate": 9.688949935231145e-08,
-      "loss": 1.7905,
+      "loss": 1.6651,
       "step": 45600
     },
     {
       "epoch": 0.914,
-      "grad_norm": 3.0062830448150635,
+      "grad_norm": 0.9774934649467468,
       "learning_rate": 9.256352338925744e-08,
-      "loss": 1.8141,
+      "loss": 1.6656,
       "step": 45700
     },
     {
       "epoch": 0.916,
-      "grad_norm": 2.7217137813568115,
+      "grad_norm": 0.900368332862854,
       "learning_rate": 8.833451861644582e-08,
-      "loss": 1.7936,
+      "loss": 1.6617,
       "step": 45800
     },
     {
       "epoch": 0.918,
-      "grad_norm": 2.7299764156341553,
+      "grad_norm": 0.9704650044441223,
       "learning_rate": 8.420265537757238e-08,
-      "loss": 1.8097,
+      "loss": 1.6744,
       "step": 45900
     },
     {
       "epoch": 0.92,
-      "grad_norm": 2.784201145172119,
+      "grad_norm": 0.9401925802230835,
       "learning_rate": 8.016810010348542e-08,
-      "loss": 1.8083,
+      "loss": 1.6708,
       "step": 46000
     },
     {
       "epoch": 0.92,
-      "eval_loss": 1.7040386199951172,
-      "eval_runtime": 24.593,
-      "eval_samples_per_second": 203.31,
-      "eval_steps_per_second": 3.212,
+      "eval_loss": 1.7872852087020874,
+      "eval_runtime": 24.9733,
+      "eval_samples_per_second": 200.213,
+      "eval_steps_per_second": 3.163,
       "step": 46000
     },
     {
       "epoch": 0.922,
-      "grad_norm": 2.5894486904144287,
+      "grad_norm": 0.9447685480117798,
       "learning_rate": 7.62310153054835e-08,
-      "loss": 1.8061,
+      "loss": 1.6587,
       "step": 46100
     },
     {
       "epoch": 0.924,
-      "grad_norm": 2.8261611461639404,
+      "grad_norm": 0.933417558670044,
       "learning_rate": 7.239155956876798e-08,
-      "loss": 1.7998,
+      "loss": 1.6658,
       "step": 46200
     },
     {
       "epoch": 0.926,
-      "grad_norm": 2.639902114868164,
+      "grad_norm": 0.9723806381225586,
       "learning_rate": 6.864988754605645e-08,
-      "loss": 1.7951,
+      "loss": 1.6625,
       "step": 46300
     },
     {
       "epoch": 0.928,
-      "grad_norm": 2.8672890663146973,
+      "grad_norm": 0.9247879981994629,
       "learning_rate": 6.500614995135185e-08,
-      "loss": 1.7955,
+      "loss": 1.6673,
       "step": 46400
     },
     {
       "epoch": 0.93,
-      "grad_norm": 2.8134496212005615,
+      "grad_norm": 0.9378394484519958,
       "learning_rate": 6.146049355387345e-08,
-      "loss": 1.7781,
+      "loss": 1.6734,
       "step": 46500
     },
     {
       "epoch": 0.932,
-      "grad_norm": 2.931941032409668,
+      "grad_norm": 0.982897937297821,
       "learning_rate": 5.8013061172144115e-08,
-      "loss": 1.7983,
+      "loss": 1.6624,
       "step": 46600
     },
     {
       "epoch": 0.934,
-      "grad_norm": 2.481424570083618,
+      "grad_norm": 0.940388023853302,
       "learning_rate": 5.466399166823733e-08,
-      "loss": 1.7801,
+      "loss": 1.6582,
       "step": 46700
     },
     {
       "epoch": 0.936,
-      "grad_norm": 2.8001270294189453,
+      "grad_norm": 0.9573161005973816,
       "learning_rate": 5.141341994218396e-08,
-      "loss": 1.8017,
+      "loss": 1.6685,
       "step": 46800
     },
     {
       "epoch": 0.938,
-      "grad_norm": 2.608473777770996,
+      "grad_norm": 0.9304656982421875,
       "learning_rate": 4.826147692653993e-08,
-      "loss": 1.8087,
+      "loss": 1.6616,
       "step": 46900
     },
     {
       "epoch": 0.94,
-      "grad_norm": 2.6505825519561768,
+      "grad_norm": 0.9191578030586243,
       "learning_rate": 4.520828958111012e-08,
-      "loss": 1.8083,
+      "loss": 1.6703,
       "step": 47000
     },
     {
       "epoch": 0.94,
-      "eval_loss": 1.6985453367233276,
-      "eval_runtime": 24.6165,
-      "eval_samples_per_second": 203.116,
-      "eval_steps_per_second": 3.209,
+      "eval_loss": 1.7930216789245605,
+      "eval_runtime": 24.945,
+      "eval_samples_per_second": 200.441,
+      "eval_steps_per_second": 3.167,
       "step": 47000
     },
     {
       "epoch": 0.942,
-      "grad_norm": 2.6877031326293945,
+      "grad_norm": 0.9425594210624695,
       "learning_rate": 4.2253980887835275e-08,
-      "loss": 1.8043,
+      "loss": 1.669,
       "step": 47100
     },
     {
       "epoch": 0.944,
-      "grad_norm": 2.619295120239258,
+      "grad_norm": 0.9412608742713928,
       "learning_rate": 3.9398669845838735e-08,
-      "loss": 1.7856,
+      "loss": 1.665,
       "step": 47200
     },
     {
       "epoch": 0.946,
-      "grad_norm": 2.698773145675659,
+      "grad_norm": 0.9414757490158081,
       "learning_rate": 3.66424714666333e-08,
-      "loss": 1.8211,
+      "loss": 1.6649,
       "step": 47300
     },
     {
       "epoch": 0.948,
-      "grad_norm": 2.674280881881714,
+      "grad_norm": 0.9912458062171936,
       "learning_rate": 3.398549676948776e-08,
-      "loss": 1.7781,
+      "loss": 1.6797,
       "step": 47400
     },
     {
       "epoch": 0.95,
-      "grad_norm": 2.6320443153381348,
+      "grad_norm": 0.9361893534660339,
       "learning_rate": 3.1427852776955424e-08,
-      "loss": 1.8048,
+      "loss": 1.6743,
       "step": 47500
     },
     {
       "epoch": 0.952,
-      "grad_norm": 2.6440446376800537,
+      "grad_norm": 0.9305044412612915,
       "learning_rate": 2.896964251056289e-08,
-      "loss": 1.7889,
+      "loss": 1.6682,
       "step": 47600
     },
     {
       "epoch": 0.954,
-      "grad_norm": 2.6774067878723145,
+      "grad_norm": 0.9433292150497437,
       "learning_rate": 2.6610964986662214e-08,
-      "loss": 1.8091,
+      "loss": 1.6582,
       "step": 47700
     },
     {
       "epoch": 0.956,
-      "grad_norm": 2.588038682937622,
+      "grad_norm": 0.9710184931755066,
       "learning_rate": 2.43519152124394e-08,
-      "loss": 1.8105,
+      "loss": 1.6702,
       "step": 47800
     },
     {
       "epoch": 0.958,
-      "grad_norm": 2.654189348220825,
+      "grad_norm": 0.9131361842155457,
       "learning_rate": 2.219258418208997e-08,
-      "loss": 1.7772,
+      "loss": 1.662,
       "step": 47900
     },
     {
       "epoch": 0.96,
-      "grad_norm": 2.593979597091675,
+      "grad_norm": 0.9314263463020325,
       "learning_rate": 2.0133058873153532e-08,
-      "loss": 1.7769,
+      "loss": 1.6642,
       "step": 48000
     },
     {
       "epoch": 0.96,
-      "eval_loss": 1.6824121475219727,
-      "eval_runtime": 24.6519,
-      "eval_samples_per_second": 202.824,
-      "eval_steps_per_second": 3.205,
+      "eval_loss": 1.8063265085220337,
+      "eval_runtime": 24.8116,
+      "eval_samples_per_second": 201.519,
+      "eval_steps_per_second": 3.184,
       "step": 48000
     },
     {
       "epoch": 0.962,
-      "grad_norm": 2.6451854705810547,
+      "grad_norm": 0.9380995035171509,
       "learning_rate": 1.8173422243008553e-08,
-      "loss": 1.799,
+      "loss": 1.6747,
       "step": 48100
     },
     {
       "epoch": 0.964,
-      "grad_norm": 2.7566401958465576,
+      "grad_norm": 0.9288477301597595,
       "learning_rate": 1.631375322553336e-08,
-      "loss": 1.7899,
+      "loss": 1.6568,
       "step": 48200
     },
     {
       "epoch": 0.966,
-      "grad_norm": 2.6575381755828857,
+      "grad_norm": 0.9180007576942444,
       "learning_rate": 1.4554126727924223e-08,
-      "loss": 1.7949,
+      "loss": 1.6617,
       "step": 48300
     },
     {
       "epoch": 0.968,
-      "grad_norm": 2.849346160888672,
+      "grad_norm": 0.966909646987915,
       "learning_rate": 1.2894613627680297e-08,
-      "loss": 1.7974,
+      "loss": 1.6668,
       "step": 48400
     },
     {
       "epoch": 0.97,
-      "grad_norm": 2.7228503227233887,
+      "grad_norm": 0.9708278775215149,
       "learning_rate": 1.1335280769746993e-08,
-      "loss": 1.8043,
+      "loss": 1.6664,
       "step": 48500
     },
     {
       "epoch": 0.972,
-      "grad_norm": 2.572453260421753,
+      "grad_norm": 0.9495335221290588,
       "learning_rate": 9.876190963824806e-09,
-      "loss": 1.7969,
+      "loss": 1.6691,
       "step": 48600
     },
     {
       "epoch": 0.974,
-      "grad_norm": 2.6655948162078857,
+      "grad_norm": 0.9480617046356201,
       "learning_rate": 8.517402981838008e-09,
-      "loss": 1.7997,
+      "loss": 1.6618,
       "step": 48700
     },
     {
       "epoch": 0.976,
-      "grad_norm": 2.4329028129577637,
+      "grad_norm": 0.9502726793289185,
       "learning_rate": 7.258971555568761e-09,
-      "loss": 1.7899,
+      "loss": 1.6725,
       "step": 48800
     },
     {
       "epoch": 0.978,
-      "grad_norm": 2.910684108734131,
+      "grad_norm": 0.9565762877464294,
       "learning_rate": 6.100947374451105e-09,
-      "loss": 1.8073,
+      "loss": 1.6624,
       "step": 48900
     },
     {
       "epoch": 0.98,
-      "grad_norm": 3.1359446048736572,
+      "grad_norm": 0.9583128094673157,
       "learning_rate": 5.043377083530365e-09,
-      "loss": 1.7929,
+      "loss": 1.6663,
       "step": 49000
     },
     {
       "epoch": 0.98,
-      "eval_loss": 1.6812490224838257,
-      "eval_runtime": 24.608,
-      "eval_samples_per_second": 203.186,
-      "eval_steps_per_second": 3.21,
+      "eval_loss": 1.8175357580184937,
+      "eval_runtime": 25.0918,
+      "eval_samples_per_second": 199.268,
+      "eval_steps_per_second": 3.148,
       "step": 49000
     },
     {
       "epoch": 0.982,
-      "grad_norm": 2.6072099208831787,
+      "grad_norm": 0.9679476022720337,
       "learning_rate": 4.0863032815838254e-09,
-      "loss": 1.7864,
+      "loss": 1.6649,
       "step": 49100
     },
     {
       "epoch": 0.984,
-      "grad_norm": 3.0630269050598145,
+      "grad_norm": 0.9047960638999939,
       "learning_rate": 3.2297645194040438e-09,
-      "loss": 1.7974,
+      "loss": 1.6691,
       "step": 49200
     },
     {
       "epoch": 0.986,
-      "grad_norm": 2.7400732040405273,
+      "grad_norm": 0.9622616767883301,
       "learning_rate": 2.4737952982481494e-09,
-      "loss": 1.8015,
+      "loss": 1.6649,
       "step": 49300
     },
     {
       "epoch": 0.988,
-      "grad_norm": 2.85768461227417,
+      "grad_norm": 0.9746965169906616,
       "learning_rate": 1.8184260684456223e-09,
-      "loss": 1.7929,
+      "loss": 1.6829,
       "step": 49400
     },
     {
       "epoch": 0.99,
-      "grad_norm": 2.7619926929473877,
+      "grad_norm": 0.9618674516677856,
       "learning_rate": 1.2636832281737176e-09,
-      "loss": 1.8059,
+      "loss": 1.6645,
       "step": 49500
     },
     {
       "epoch": 0.992,
-      "grad_norm": 2.5949032306671143,
+      "grad_norm": 0.9379156827926636,
       "learning_rate": 8.095891223933172e-10,
-      "loss": 1.7836,
+      "loss": 1.6675,
       "step": 49600
     },
     {
       "epoch": 0.994,
-      "grad_norm": 2.563429832458496,
+      "grad_norm": 1.0468147993087769,
       "learning_rate": 4.5616204194881554e-10,
-      "loss": 1.8171,
+      "loss": 1.6665,
       "step": 49700
     },
     {
       "epoch": 0.996,
-      "grad_norm": 2.933175563812256,
+      "grad_norm": 0.9367111921310425,
       "learning_rate": 2.034162228309322e-10,
-      "loss": 1.8067,
+      "loss": 1.6605,
       "step": 49800
     },
     {
       "epoch": 0.998,
-      "grad_norm": 2.878960132598877,
+      "grad_norm": 0.9311544895172119,
       "learning_rate": 5.1361845604946944e-11,
-      "loss": 1.8001,
+      "loss": 1.6604,
       "step": 49900
     },
     {
       "epoch": 1.0,
-      "grad_norm": 2.752953290939331,
+      "grad_norm": 0.9327343106269836,
       "learning_rate": 5.034998806774383e-15,
-      "loss": 1.7611,
+      "loss": 1.6704,
       "step": 50000
     },
     {
       "epoch": 1.0,
-      "eval_loss": 1.7131491899490356,
-      "eval_runtime": 24.5919,
-      "eval_samples_per_second": 203.319,
-      "eval_steps_per_second": 3.212,
+      "eval_loss": 1.813862681388855,
+      "eval_runtime": 25.1017,
+      "eval_samples_per_second": 199.19,
+      "eval_steps_per_second": 3.147,
       "step": 50000
     }
   ],